TELLES > Oporto, Portugal > Firm Profile

Av. Marechal Gomes da Costa, 1131
4150-360 Oporto

Portugal > Energy and natural resources Tier 2

TELLES has broad capabilities across the entire spectrum of mandates pertaining to the energy sector and carbon markets, and particularly excels in M&A and project finance. The department is adept in acting for high-profile public and private-sector clients locally and abroad in relation to licensing, regulatory issues, public law, environmental and climate change-related matters. Renewables is an area of strength for the team, which is jointly led by Bruno Azevedo Rodrigues and Ivone Rocha, the latter of whom splits her time between Lisbon and Porto. Rocha excels in the development of energy production projects and energy efficiency, while Azevedo Rodrigues takes the helm in financing work. Mariana Figueiredo is also part of the core team and is noted for her strong background in energy and environmental law.

Practice head(s):

Bruno Azevedo Rodrigues; Ivone Rocha

Other key lawyers:

Mariana Figueiredo; João Lupi


‘Telles proved to be very prepared to respond to the challenges and questions facing companies in the energy and natural resources sector. It stands out from most law firms for its client service, availability and responsiveness. The energy and corporate departments are excellent, staffed by lawyers with deep knowledge and experience in these areas.’

‘Ivone Rocha proved to be an excellent team leader and interlocutor between the client and the firm. She has great knowledge in the field of Energy.’

‘Mariana Figueiredo is also recommended for her availability, client service and for her comprehensive and in-depth knowledge of the energy sector.’

‘Mariana Figueiredo provides full monitoring of all processes, even if they may be from different areas.’

‘The teams we have worked with so far have been fantastic and we are extremely satisfied.’

‘The Telles Energy and Natural Resources Team stands out for its extensive knowledge of the sector at national level, for its ability to internally coordinate various areas, providing the client with a single point of contact, and also for the dynamism and initiative with which it analyses the various themes and proposes solutions. I recommend Mariana Figueiredo, namely for her know-how in the energy sector, for the knowledge and contacts she not only has in the national and international scene, as well as the availability and professionalism with which she follows the various topics under analysis.’

‘We worked with Mariana Figuerido and Ivone Rocha. They provided very clear feedback and advice, clear reports.’

‘We value Mariana Figueiredo’s efforts to go above and beyond, and we’ve found that it’s not just her ability to provide us with competent advice, but her communication and collaboration skills that make her noteworthy.’

Key clients












































Work highlights

  • Advising LIPOR on the design of a new and alternative legal and contractual remuneration model for its waste-to-energy plant.
  • Acting for a consortium in the licensing and negotiation with the Portuguese government and local authorities of a hydrogen electrolyser and 350MW solar PV plants.
  • Assisting AdEPORTO in the design and  implementation of the PEER Project.

Portugal > Planning law Tier 2

TELLES' planning practice is praised by clients for its urban development expertise, and is particularly strong in development, construction and licensing instructions. In Porto, Pedro Almeida e Sousa's extensive and longstanding real estate experience is valued by clients, particularly in historic regeneration projects. Miguel Nogueira Leite is based in Lisbon, from where he co-leads the team; he has a broad practice across the retail, tourism, agricultural and industrial sectors. Senior associate Frederico Styliano Reis and associate Bernardo Gomes De Castro are further names to note.

Practice head(s):

Pedro Almeida e Sousa; Miguel Nogueira Leite

Other key lawyers:

Frederico Styliano Reis; Bernardo Gomes de Castro


‘TELLES offer a great relationship with the client, involvement in the themes with great collaboration and a good knowledge in urbanism matters.’

‘The firm has deep knowledge in matters of urbanism, with specialists in the area are highlights for this law firm, resulting in a quality of service provided above what is normal in the market.’

Key clients
















Work highlights

  • Advised RAR Group on the licensing of the rehabilitation of Quinta de S. José de Ribamar (a 16th Century monastery and palace) for the development of a residential area.
  • Advised Lousas de Valondo S.A on the preparation of a Prior Information Request and a development contract for the maintenance of the construction capacity of land located in Valongo.
  • Advised SIS Groupe on the licensing process of an operation to rehabilitate a city block in ruins in the historical centre of Braga and  build an assisted living and care centre.

Portugal > Commercial, corporate and M&A Tier 3

The ‘highly professional team' at TELLES is ‘technically well prepared' and possesses ‘vast transactional experience'. With a multidisciplinary approach bringing together expertise in tax, competition, finance, real estate and labour, the practice is well versed across the gamut of transactional work and attracts instructions from multinational corporations, venture capital firms, private equity houses and domestic clients across the insurance, industrial, healthcare, telecoms, infrastructure and construction sectors. The team is jointly led by Carlos Lucena, who has a strong track record in major M&A deals involving the northern region of Portugal, and Francisco Espregueira Mendes, who has combined knowledge of corporate, commercial and competition issues.

Practice head(s):

Carlos Lucena; Francisco Espregueira Mendes

Other key lawyers:

Bruno Azevedo Rodrigues; Miguel Carvalho; Cristina Ferreira; Mariana Ferreira Martins; Mariana Vilaça Fernandes; Diogo Damião


‘Hands-on team, very flexible in an unusual M&A environment, high availability and solutions-driven to make the acquisition happen.’

‘Multidisciplinary team with great experience, dynamic and proactive.’

We have worked with Diogo Damião and we were very happy with his work. Very detailed and quality-oriented, available 24/7. Not only is he available, but also very attentive, careful, detailed and result-oriented.’

‘Highly professional team, that is technically well prepared, with vast transactional experience, and above all available to the client, at any time.’

‘Although a large part of the process was done online, the rapport that was created between both work teams was notorious, making clear Telles’ genuine concern for everyone involved. The two prominent figures in this process are Francisco Espregueira Mendes and Diogo Damião.’

‘Diogo Damião has an extraordinary capacity for work and a tireless focus on the strict defence of his client’s rights. He helped us to clearly see the weaknesses of our contract and led the conversation to close the door to the possibility of future hardships as much as possible.’

‘Francisco Espregueira Mendes: the accumulated experience he puts at the service of the process and his team, with an unparalleled pragmatism, fully knowing which battles we have to fight and which others we can give in favour of the other party.’

‘Ability to execute and a vision oriented towards solving problems and leading to the conclusion of deals.’

Key clients























































Work highlights

  • Advised METALOGALVA on the acquisition of the lighting division of the Europoles Group in Germany, Poland and France.
  • Advised PLURIS INVESTMENTS on the direct acquisition of 30% of Media Capital from the Spanish group PRISA, and in the public offering for the acquisition of the remaining share capital.
  • Assisted VALLIS CAPITAL PARTNERS in the sale of IMPERIAL, the largest Portuguese chocolate producer, to Spain-based Chocolates Valor.

Portugal > Projects and project finance Tier 3

A range of clients from the banking, infrastructure, tourism, real estate, transport, pharmaceutical and distribution sectors turn to the projects team at TELLES for advice on project finance, licensing issues, project agreements and M&A transactions involving assets in regulated industries. Department head Bruno Azevedo Rodrigues is highly regarded for his experience in the energy sector, and primarily assists with project finance and secured lending in relation to infrastructure projects.

Practice head(s):

Bruno Azevedo Rodrigues

Other key lawyers:

Delber Pinto Gomes; João Lupi


‘The responsiveness and quality of legal services in the energy sector are exemplary. The in-depth knowledge of the energy sector – and its agents – translates into an enormous asset in the legal opinion that is given to the client.’

‘Flexibility and availability. Bruno Azevedo is recommended.’

Key clients













































Work highlights

  • Advising the consortium comprised of Island Renewable Ltd and CJR Group in relation to five projects of new solar PV power plants to be located near the Municipality of Sine.
  • Advising the consortium comprised of MAN, WINPOWER and OBERON INFRASTRUCTURES ADVISORS in the licensing and negotiation with the Portuguese government and local authorities of a hydrogen electrolyser and 350MW solar PV plants.
  • Acting for MIROVA in relation to the acquisition of six hydroelectric plants from EDP.

Portugal > Tax Tier 3

With a solid on-the-ground presence in Lisbon and Porto, the tax department at TELLES is well placed to handle work for a diverse range of blue-chip companies, banks and family offices in relation to cross-border business restructuring, M&A, private equity deals, and international wealth management. Tax litigation is another area of focus for the group. Miguel Torres is ‘an outstanding lawyer' who combines expertise in corporate and tax law. João Magalhães Ramalho is well versed in domestic and cross-border M&A, and real estate tax cases. ‘With a very solid background in international taxation', João Luís Araújo is trusted by corporate and private clients. André Gonçalves is also noted for international and national tax issues.

Practice head(s):

Miguel Torres; João Magalhães Ramalho; João Luís Araújo; André Gonçalves

Other key lawyers:

Abílio Rodrigues


‘Telles’ tax practice combines expertise (both in terms of knowledge and experience) with a high level of dedication to clients and their needs. Though part of an independent law firm, you can tell the practice is truly international and some of its partners have been part of large multinational organisations in previous roles.’

‘João Luís Araujo is a tax partner with a very solid background in international taxation, having served in various roles before Telles. João is a quick thinker, responsive to requests and focused on delivering results.’

‘Miguel Torres is an outstanding lawyer with great communication skills’.

‘Miguel Torres- in addition to his technical skills, I highlight his proactivity and availability.’

‘Competence, professionalism and availability.’

‘I have found the team at Telles to operate an extremely professional service, dealing expertly with ultra-high-net-worth international clients – an outstanding firm in the Portuguese market.’

‘João Luís Araújo is an exceptional lawyer, technically excellent and very personable with great client skills. I would have no hesitation in recommending him to any of my clients needing Portuguese tax and estate planning advice.’

Key clients










Grupo NORS

Quinta do Crasto


Garcia Garcia





Grupo Vigent

Família Jervell

Atlantic SGFII



Grupo Oito Em ponto

Grupo BRS

Grupo Planalto

Work highlights

  • Advised Vallis on tax matters during the process of sale of its participation in Imperial (the biggest chocolate producer in Portugal).
  • Assisted the partners of Insparya Group in the restructuring of Grupo Saúde Viável.

Portugal > Dispute resolution Tier 4

Singled out for its strong presence in Porto, TELLES' dispute resolution group is retained by clients from a multitude of sectors in arbitration and litigation cases. The team is prepared to assist with domestic and cross-border contentious matters, and primarily focuses on civil and commercial disputes, notably those arising from shareholder and corporate issues, as well as conflicts in relation to M&A deals, insurance, construction and distribution agreements. André Navarro de Noronha ‘takes the lead in litigation', while Fernando Pizarro Monteiro is recommended for restructuring and insolvency cases. Nuno Miguel Lourenço is the name to note in Lisbon for commercial and civil litigation.

Practice head(s):

André Navarro de Noronha; Fernando Pizarro Monteiro; Nuno Miguel Lourenço

Other key lawyers:

Maria João Meireles; Catarina Alegre


‘Excellent monitoring, proactive team and good advice.’

‘I recommend André Navarro Noronha for his technical excellence and experience; Catarina Alegre for her work capacity, study of processes, and advice that is focused on the real needs of the client.’

‘A great deal of work was needed to survey the questions listed by the other party and to answer them, which was given in a very complete, exhaustive way, completely destroying the opposing party’s arguments, based on a very complete explanation of the issues.’

‘The interactions with Maria João Meireles were extremely fruitful, showing not only a great availability to listen to the client, but also a great concern to “assemble” a case that was unassailable by the other party. It is clearly felt that she is an attentive and very methodical lawyer, with a great sensitivity to the issues in question and deep technical and legal knowledge. As for Fernando Pizarro, he has a perfect command of the issues in question and performs the necessary monitoring of his team.’

‘Telles’ litigation team demonstrates a strong response capacity and is particularly recognised in the Porto market. André Navarro de Noronha has demonstrated that he takes the lead in the litigation processes entrusted to him.’

‘Telles is a unique law firm in the context of the Portuguese market, especially due to the balance it has between the virtues of a large firm (level and transversality of knowledge, professionalism, compliance, networking, etc.) and those of a small firm (service very personalized, availability, high client service, flexibility, availability, etc.).’

‘The team demonstrated that they had good knowledge of the matters in which they were consulted. It was, therefore, able to provide quality collaboration and it lived up to our expectations. We would particularly highlight Maria João Meireles and Fernando Pizarro.’

Key clients

Körber Supply Chain PT, S.A.

Portugal > Employment Tier 4

TELLES' employment practice is highly specialised in labour and social security law, advising clients across a wide range of employment matters including contracts, remuneration, pension plans and restructuring protocols. In addition, the firm also advises clients on day-to-day matters. Gonçalo Pinto Ferreira leads the practice, and has over 20 years of experience in advising Portuguese and international clients, often focusing on large-scale restructurings.

Practice head(s):

Gonçalo Pinto Ferreira

Other key lawyers:

Sofia Pamplona; João Dotti de Carvalho


‘Very good knowledge of our company, sector, global market. Very good availability. Clear formalized opinion that protects the interest of the company but at the same time adapted to our operational constraints.’

‘This is a team of very professional lawyers who can give a quick and quality response. They have solid and extensive technical knowledge to be able to respond to clients. They are very accessible and adapt the response to the necessities of clients.’

‘The relationship with this law firm is, in fact, excellent because the established relationship of trust is based on the quality and speed of the services provided.’

‘The service provided by Telles and its team led by Gonçalo Pinto Ferreira provide an excellent service that excels in the ability to combine legal frameworks with the operational vision of the company’s activity.’

‘They act as partners. We can find in depth all the details as well as work history which helps us and streamlines in the decision-making process as well as in the preparation of legal parts.’

‘Sofia Pamplona is a very dedicated and available lawyer, providing high quality advice on Employment law, with accurate responses perfectly suited to her clients’ needs.’

Key clients

Leroy Merlin

Pepsico Group

Tyco Eletronics (TE)


Accor Hotels Group

GKN Automotive

Group Sonae

PSA Sines

Group Nors

Somincor (Group Lundin Mining)

Work highlights

  • Assists Accor Group, owner of the hotels Sofitel, Novotel, Mercure and Ibis, in all employment-related matters, impacting around 40 hotels and 600 employees.
  • Assisting PepsiCo Group in Portugal, its marketing and distribution company (Matudis) and the respective manufacturing company (Matutano), covering all employment-related matters.
  • Advising Leroy Merlin on a daily basis on all labour issues, including the reorganisation of working patterns and store schedules.

Portugal > EU and competition Tier 4

Highlighted for its ‘professionalism, availability and support', TELLES' competition department acts for multinational corporations, foreign subsidiaries and Portuguese companies in an array of matters. Recent instructions showcase the team’s strength in handling merger filings, anti-competitive practices, state aid and abuse of dominance cases. The team is jointly led by Francisco Espregueira Mendes and Leyre Prieto, the latter of whom is particularly active in merger control proceedings before the PCA and private enforcement cases. Of counsel Joaquim Caimoto Duarte, who joined the team in May 2020 from Uría Menéndez – Proença Carvalho, is noted for his wealth of experience in cross-border merger control and cartel cases.

Practice head(s):

Francisco Espregueira Mendes; Leyre Prieto

Other key lawyers:


‘Excellent command of the area of ​​reference and close contact with the client.’

‘Joaquim Caimoto Duarte has excellent mastery of his area of ​​expertise. He is extremely experienced and knowledgeable. Availability to the client.’

‘Telles is a unique law firm in the context of the Portuguese market, especially due to the balance it has between the virtues of a large firm (level and transversality of knowledge, professionalism, compliance, networking, etc.) and those of a small firm (personalised service, availability, flexibility, availability, etc.). I would like to highlight Telles’ ability and speed to adapt to remote work in the face of adversity caused by the Covid-19 pandemic, which did not negatively affect Telles’ performance and delivery.’

‘All elements of the Telles team in this area of ​​expertise are above average. I recommend Leyre Prieto especially for her level of knowledge in the area and high availability.’

‘Professionalism, availability and support.’

Key clients



MATUDIS – Comércio de Produtos Alimentares, Lda.

INTER-RISCO –Sociedade de Capital de Risco, S.A.

Tintas Robbialac, S.A.

Super Bock Bebidas, S.A.



RNM – Investimentos e Gestão, S.A.

Vallis Capital Partners, SGPS, S.A.

Fashion Division, S.A.

AMC Natural Drinks SL

Pluris Investments, S.A.

Crédito Y Caución, S.A. de Seguros Y Reaseguros – Sucursal em Portugal

Descours & Cabaud



Vortal SGPS

Portugal > Data privacy and data protection

Under the leadership of Lisbon-based partner Pedro Vidigal Monteiro, TELLES is instructed by an impressive corporate client base from a variety of industries seeking advice on data protection compliance programmes, GDPR issues and data incidents.

Portugal > Public law

With a strong presence in the north of Portugal, TELLES offers advice on administrative law and public procurement to a number of public entities; Porto-based partner Pedro Almeida e Sousa oversees the department.

Portugal > Real estate and construction

TELLES' real estate practice sits within the firm's wider real estate, construction and urban planning department, offering a multidisciplinary approach to provide advice to private and public international and domestic clients; the practice is co-led by Pedro Almeida e Sousa in Porto and Miguel Nogueira Leite in Lisbon.

Portugal > TMT

TELLES acts for private and public entities, including software and technology companies, in a range of data protection and TMT matters, involving mobility, artificial intelligence, internet of things and blockchain issues; Pedro Vidigal Monteiro leads the digital, privacy and cybersecurity department.

TELLES, widely recognised for providing a high standard of legal advice, is celebrating 30 years as independent portuguese law firm and the history that brought us to this point.

Since 1936, the firm’s lawyers and staff have worked across industries, practices and jurisdictions to provide clients with simple, actionable approaches to their most complex challenges. TELLES’s ADN still the same: delivery legal work with highest standard.

Located in Lisbon and Porto, TELLES is offering its clients, from major multinationals and mid-caps to enterprising start-ups and private clients, namely the higth net worth private clients, much more than just the law. In a fast-changing world, the 120 lawyers of the firm are committed to worked closely with clients to help them face their new and complex challenges, always with the personal involvement of its partners.

TELLES is a full-service law firm, composed with expert lawyers that having worked on most of the major transactions occurred on the portuguese market, representing international and national clients on their most strategic domestic and cross-border transactions.

Follow us on LinkedIn

Commercial, corporate and M&A
Litigation & Arbitration
Employment & Social Security
Environment & Climate Change
Energy & Natural Resources
Finance, Projects & Capital Markets
Competition & Antitrust
Digital, Privacy & Cybersecurity
Intellectual Property
Restructuring and Insolvency
Public Law & Public Procurement
Real Estate & Urban Planning
Fernando Pizarro Monteiro photo Fernando Pizarro MonteiroPartner
Joana Telles de Abreu photo Joana Telles de AbreuPartner
Miguel Torres photo Miguel TorresManaging Partner and Head of Tax
Number of lawyers : 120
Other offices : Oporto

TELLES is a recommended firm in the Green Guide EMEA 2022. Learn more about this recognition.

Hydrogen: present challenges of the energy source of the future


Energy transition will be a significant part of climate action for the next decades, and a fundamental instrument for developed economies – which have mostly committed to carbon neutrality at or around 2050 – to achieve their goals and commitments towards a more sustainable future. It poses significant technical, regulatory, financial, and legislative challenges yet to be fully answered.

One of the potential tools in the energy transition toolbox is Green Hydrogen: it takes up surplus renewable power from the grid and can offer long-term storage and / or cross-sector energy delivery; it facilitates greater grid management flexibility; it can directly impact the decarbonization of hard-to-abate sectors such as heavy industry and transportation.

Policy movements towards implementing a Green Hydrogen economy in Europe are relatively recent. Many questions remain on why and how to implement Green Hydrogen projects. Analysing them from a mostly legal standpoint, we will offer some takes on what still needs to be done, and what has already taken place to secure investment interest in this nascent sector.

Part 1: Why hydrogen?

Hydrogen is not new. It is the simplest and most abundant element in the universe; its properties are well known and studied. We have used it extensively for a long period of time: as a reactant in chemical plants, as fuel in specific applications. It shall be borne in mind that global demand for Hydrogen in 2020 was 90 Mt1. Most middle schoolers in the developed world have seen the electrolysis of water as a simple, easily displayed and viewable electrochemical reaction.

Why, then, the renewed interest on hydrogen?

Its material characteristics have long been touted as one, if not the, stabilizing solution to the energy transition. In its Net Zero by 2050: A Roadmap for the Global Energy Sector report, the IEA indicates that 10% of the total final energy consumption by 2050 will be fulfilled by Green Hydrogen. It is energy-dense, easy to transport, has a wide range of potential uses from energy storage and transport, to the decarbonization of hard-to-abate sectors, where it is seen as the key for energy transition. It may also present significant advantages to other means of energy interconnection, allowing for the transport of significant amounts of renewable energy from where it is cheaply and easily generated – i.e., within the global solar belt – to other, less naturally prone locations.

Additionally, the ever-diminishing LCOE of renewable power generation, especially in the 2010-2020 decade, combined with carbon tax and credit schemes, have led us to the moment we find ourselves in: nearing price-parity between natural gas and green hydrogen. Current technologies and market practises are quoting the levelized cost of hydrogen for non-renewable hydrogen (i.e., steam methane reforming) from USD 0.5 to USD 1.7; low-carbon hydrogen (i.e. using CCUS technology) has been quoted at around USD 1 to USD 2; and Green hydrogen at USD 3 to USD 8 per kg2, with innovative project developers promising a much lower price point for their levelized cost of hydrogen, at par with natural gas.

It does, however, bring about its own challenges. Rapidly deploying massive hydrogen generation projects brings along administrative strain on regulating States and demands a legal, economic and tariff infrastructure that simply does not currently exist. Whilst some European Member States have been spearheading efforts for the expedited deployment of utility-scale generation projects, both by passing legislation and regulation to that effect3, there seems to be little consensus on how to achieve a breakthrough moment in regulation, allowing for the safe and efficient deployment of all parts of the value chain – all of them currently lacking capacity for the needs of a 2050 net zero energy system.

The recent proposals from the European Commission contained in the Hydrogen and Decarbonised Gas Package have made a shy, although welcomed, step towards an investor-friendly regulatory framework aimed at bridging the funding gap of Green hydrogen. The forethought given to dedicated hydrogen networks and regulation, the increased sector coupling in EU-wide network planning and proposing to align all EU Member States with first-movers’ efforts to adapt the auditable and well-established certification scheme already known from the 2nd Renewable Energy Directive4, all show great promise in kickstarting private deployment of these projects.

Recent developments in the European energy market have deepened the need for Green Hydrogen5. The current state of European energy markets – with great dependence from third-party states to provide baseload and peaking capacity in power markets, as a result from both the dependency on imported natural gas and policy decisions to expedite nuclear baseload plants going offline -, poses an additional strain on the already existing concerns of the security of supply. The European Commission sees Green and low-carbon Hydrogen (and other gaseous renewable and low-carbon fuels) as one of the tools to mitigate this exposure risk – and Member States should take heed and promote the reinforcement of renewable hydrogen generation capacity.

Green Hydrogen does not therefore come without its challenges. We will try to address some of these challenges from a legal point of view.

Part 2: The woes and pains of first-movers, or administrative licensing for hydrogen projects

Traditional, non-renewable, hydrogen generation is based on a well-known and established steam methane reforming, which poses no great challenge to legal and administrative systems in what regards their licensing: it is a simple industrial endeavour, which is subject to both the regulation of economic activities and environmental regulations.

Low-carbon hydrogen also poses little to no issue: the implementation of CCUS technology in steam methane reforming plants is not a legally innovative process; the techniques, technologies and their risks are known and established, and economic actors have knowledge on how they can proceed.
Green hydrogen generation, especially when obtained through the electrolysis of water using renewable power, however, has shown to be particularly troublesome for regulators and administrative entities alike.

Whilst industrial regulation and legislation should show itself to be flexible enough to allow for innovative products and processes to be deployed within their legal system, complexities can be found in the unending tendency of environmental legislation to adopt strict taxonomy systems for regulation. Taxonomy – especially when enshrined into law – has a significant disadvantage in regulating highly innovative and ever-evolving fields of industrial endeavour, and the well-known in tech phenomena of “law lags behind” is showing, once again, in the energy sector.

Diving into a specific European legal regime as an example: the Directive on Industrial Emissions, or the Integrated Pollution Prevention and Control Directive6. This legal system, which has to be mandatorily transposed by all Member States, intends on laying down “rules on integrated prevention and control of pollution arising from industrial activities” and also “designed to prevent or, where that is not practicable, to reduce emissions into air, water and land and to prevent the generation of waste, in order to achieve a high level of protection of the environment taken as a whole”. A noble and desirable goal in itself, it opted on taxonomizing all industrial activities subject to it and only exempting “research activities, development activities or the testing of new products and processes”. As a consequence of said taxonomy, the “production of inorganic chemicals, such as (…) hydrogen7, whenever “on an industrial scale”, is always subject to the Directive on Industrial Emissions.

Imposing compliance with this Directive (and its legal transposition into national legal frameworks) means demanding additional licensing and reporting obligations: from obtaining an environmental permit to operate, to proving, inter alia, that best available techniques are applied in the design, deployment and operation of the industrial endeavour, and subjecting them to emission limits for polluting substances.

By its sheer definition, the electrolysis of water using renewable power does not seem to entail any emissions of any polluting substance whatsoever: generating the necessary power has no emissions, and the only by-product is oxygen, with the industrial installation being, by design, incapable of generating any other substances. Any emissions-related regulation has therefore no purpose or fit in a simple electrolyser facility; and applying it is a prime example of the concept of deadweight loss of regulation.

The most simplified investor handbook prepared by administrative authorities in regulators currently applicable in the European Union economic area regarding this matter was issued by the Portuguese Directorate-General for Energy and Geology and the Portuguese Environmental Agency8, where the applicability of the Directive on Industrial Emissions is expressly recognized for all projects. Other environmental agencies across Europe are taking a similar stance on the issue.

Whilst merely illustrative, this example allows us to shine a light on how legal systems may struggle with innovation, especially when towards a greener energy system as a whole. Legislators and regulators shall take into account the specificities of regulating innovation and make haste in avoiding taxonomy as the panacea for all ills and pains they find in their day-to-day activity. States must quickly adopt innovation-friendly frameworks which allow for the 2nd Renewable Energy Directive’s target of a two-year full licensing period for power plants to be also applicable to utility-scale hydrogen projects.

The proposals from the European Commission contained in the Hydrogen and Decarbonised Gas Package do not significantly contribute to this regard – and it is easily understandable, given the complex regulatory process typically adopted by the European Commission. The field of innovation regulation has been starting to pick up steam in European legal scholarship, and while it still feels nascent, it can provide key takeaways on how to provide a stable and future-proof regulation which withstands innovation, simultaneously providing for economic efficiency and other public policy goals.

Part 3: Paradigm shifting, or sector coupling as the answer for an under asked question

A high renewable energy source penetration in the power mix demands innovative measures to ensure grid stability and quality, in light of the innate variability of these energy sources. This has been touted as the main cost-driver and challenge to network operators, system planners and regulators alike: how to reach a balance between cost-effectiveness of measures and quality of supply?

We believe even here Green Hydrogen is able to earn a place in the toolkit those market actors can refer to in order to tackle those challenges.

From a technical standpoint, the added long-term storage flexibility provided by gaseous fuels, such as Hydrogen, and the multiplicity of potential off takers in different economic sectors and value chains, bring with them significant added value to Hydrogen as a product. Additionally, as consumer-response mechanisms are integrated into energy markets across Europe, electrolysis shows great potential at quickly and efficiently participating in those markets, being flexible enough to expeditiously answer curtailment commands from grid operators.

Electrolysers are also at the border between power and gas markets. They are showing to be potential accelerators of sector coupling, with the multiplicity of Power-to-X solutions currently being designed and implemented. Because of that, the European Commission and other stakeholders have taken definite steps towards facilitating market entrance of these equipment.

In the aforementioned Hydrogen and Decarbonised Gas Package as presented by the European Commission, we’ve seen progress towards integration, with a proposal to unify and further take into consideration all grids – power, natural gas and dedicated hydrogen, if existing – when designing and implementing the European Ten-Year Network Development Plan. Additionally, the original European proposal for a revised Regulation on the guidelines for trans-European energy infrastructure (the TEN-E Regulation) made strides in its final version as approved by the Council, allowing both the coupling of the TEN-E and TEN-T (trans-European transport infrastructure) networks and the implementation of projects of common European interest in a decentralized, heavily coupled way, breaking new ground in allowing transnational projects to go farther than the traditional pipeline and interconnection lines paradigm. Those decisions are fully in line with the principles already enshrined and in force in the 2nd Renewable Energy Directive and the European Power Markets legislative revision of 20199.

Independent regulators have heard the call and are adjusting tariff and other regulation. One of the immediate challenges promoters face in modelling on-grid electrolysers, in order to capture surplus power and provide network services to grid operators, are grid connection and operation tariff systems capable of meeting the financial needs of this type of projects. Specifically, in the Portuguese market, ERSE10 has approved specified tariff mechanisms for standalone storage facilities, including electrolysers coupled with stationary fuel cells, which through granting several exemptions level the playing field, encourage private investment and reduce or ultimately eliminate discrimination.

Furthermore, legislative commands given by European legislators, in a direct manifestation of the energy efficiency first principle and as a contribution for the economic efficiency of grid planning exercises, make new grid capacity essentially a last-resort investment, instead favouring investments in short and long-term storage, capacity markets and other mechanisms for a flexible and functioning grid. Legislators took those instructions to heart and are enshrining those instruments in law – even though some relevant markets, such as Portugal11, Germany and the Netherlands, do not seem to have transposed the relevant European legal instruments as of yet.

Sector coupling has a relevant contribution towards the viability of Green Hydrogen projects, by multiplying the markets in which generators can act, diversifying products and / or offtakers, providing more investment safety for all participants through increased reliability and market liquidity – and legislators and regulators alike seem to have picked up their cue and are implementing definite measures to encourage investment and private party interest in these markets. The outlook appears favourable, with political statements and engagement from all stakeholders providing comfort to this strategy.

Part 4: Bridging the gap, or what support schemes and mechanisms are being sought

The novelty of massified hydrogen markets combined with the innate capital intensity of energy sector investments requires significant commitment by private parties to become first-movers in the Green Hydrogen subsector. Thus, the European Commission has tasked Member States with developing a healthy and competitive hydrogen market.

How can States and other actors improve effectiveness and attractiveness of this nascent market to private investors? As previously mentioned, cost-parity outlooks are favourable to Green Hydrogen, which has a relatively stable cost structure, when compared to both non-renewable hydrogen and to natural gas; but how do we make investors whole considering the current price gap between Green Hydrogen and the energy sources it seeks to displace?

The investment profile in the Hydrogen submarket is not dissimilar to other, renewable energy markets: significant upfront CAPEX investment and a relatively long-life equipment. The specific identifiable risks are related to the novelty of the market: will offtakers be there in the mid to long-term of electrolysers, will these offtakers pass the credit rating test, will there be robust medium to long term offtake contracts or a sufficiently liquid commodity markets for Green Hydrogen, or will investors end up with stranded assets or even worse: projects that are not bankable and will never go beyond well intended drawing boards?

Traditional CAPEX support mechanisms may not be enough. While the EU and Member States alike have been offering such support mechanisms – often including subsidies for the upcoming industry – 12, investors have been calling out for diverse public support to bridge the funding gap and decrease the natural risk exposure for these kinds of investments.

Portugal has prepared its national legislation to offer lifecycle support to Green Hydrogen projects and laid the groundwork for these to mature. The Portuguese National Hydrogen Strategy plans to offer supply-and-demand CFD auctions, targeted at hydrogen consumers and / or Natural Gas retail sellers, whereas against a firm purchase agreement with a hydrogen producer the national Environmental Fund will co-pay for the difference between commodity-market prices for the fuel they are substituting and the agreed price, on a euro for euro basis. These mechanisms will allow for offtake stability and provide cost-neutrality for end-users in an economically efficient way. The legislation is equally designed for other, innovative financing solutions to be quickly deployed.

EU-wide and States’ aid measures are in force and are expected to be reinforced in the coming years. As the European Commission has often stated, the Green Hydrogen market needs to be built from the ground up – and offering a firm offtake guarantee will allow for greater confidence and bankability for investors and capital markets at large.

In conclusion

Hydrogen is likely to become one very significant part of the answer to the challenges posed by energy transition, given its specific characteristics and despite the identified hardships, simply because there is no current viable alternative to tackle the pressing climate change challenges and to reach Europe’s climate binding targets. As we all feel and see the need for the pace of energy transition to pick up, all involved stakeholders should make strides at taking action on what needs to be changed in legal, financial and regulatory frameworks in order for these projects to be able to take off.

Not everything is done – but some legwork has already taken place. Whilst some legacy regulatory challenges remain and still require attention – even though, it shall not be forgotten, that the current scenario may also provide some welcomed regulatory flexibility for first-movers that may present some opportunities -, investors and promoters alike should take confidence and comfort from constant and positive political attention and will to develop a Green Hydrogen economy. The Fit for 55 renewed ambition for the decarbonization of European economies will take an all-hands-on-deck approach to be fulfilled – and public stakeholders seem to be taking that approach.


1. Source: IEA (2021) Global Hydrogen Review 2021, Revised version (November 2021), All rights reserved.
2. Source: IEA (2021) Global Hydrogen Review 2021, Revised version (November 2021), All rights reserved.
3. With the notable effort Portugal has made in passing both Decree-law no. 60/2020, of the 17th of august, establishing a Guarantee of Origin scheme for renewable and low-carbon gases, and Decree-law no. 62/2020, of the 28th of august, regulating the deployment of renewable and low-carbon gas projects and their mix into the natural gas grids, as the first steps in deploying the National Hydrogen Strategy.
4. Directive (EU) 2018/2001 of the European Parliament and of the Council of 11 December 2018 on the promotion of the use of energy from renewable sources
5. Note from the authors: we opted to refer, for the Reader’s convenience, to Renewable Hydrogen as “Green Hydrogen”, and to any other means of generating low-carbon hydrogen, such as using the output of CCUS natural gas power plants and/or nuclear power plants as “low-carbon Hydrogen”. The use of this nomenclature should not be construed as the opinion of the authors in the ongoing policy debate on whether and how to discriminate between low-carbon power sources for Hydrogen generation.
6. Directive 2010/75/EU of the European Parliament and of the Council, of 24 November 2010, as amended.
7. Annex I, no. 4.1(a) of Directive 2010/75/EU, as amended.
8. DGEG and APA, 2021, Guia do Promotor “Legislação e regulação para a Economia do Hidrogénio”, consulted at on 31.12.2021 (in Portuguese).
9. Mainly comprised of Regulation 2019/943 and Directive 2019/944.
10. The independent regulator for both the power and natural gas grids.
11. A revision of the power sector legislation, containing the transposition of Directive 2019/944 and which takes into account Regulation 2019/943 has been approved by the Portuguese Government but is not, at the moment of writing, published and in force.
12. Namely the European Commission’s Hydrogen Public Funding Compass, and Portugal’s 2020 PO SEUR (under Horizon 2020) and Recovery and Resilience Plan support tenders.


Contributed by Bruno Azevedo Rodrigues, Ivone Rocha, Mariana Figueiredo and Rui Ferreira de Almeida of TELLES.



According to EUROSTAT, more people started to work from home following the introduction of the social distancing measures in response to the COVID-19 pandemic. In 2020, 12% of employed people aged 20-64 in the EU usually worked from home, while this share had remained constant at around 5 or 6% over the past decade.

Around one in four employed usually worked from home in these capital regions: Eastern and Midland in Ireland (25%), Wien in Austria and Hovedstaden in Denmark (both 24%) as well as Île-de-France in France, Utrecht in the Netherlands, Luxembourg (single region) and Lisboa in Portugal (all 23%).

In the context of this increasingly importance of remote work, Portugal approved new employment legislation aimed to rule the relationship between employer and employee within this new model of work organization. Effective as of January 1, 2022, new legislation is in place addressing not only the remote work, but also elaborating on the employee’s right to disconnect.

Despite some new challenges, Portugal remains focused in getting the conditions for wider implementation of flexible ways of work. This is an opportunity also for everyone, including for newcomers, to introduce harmonized policies. A good set of policies on flex work and the right to disconnect will be helpful to meet the requirements of Portuguese law and to bring labour relations to a new level.

Working from Home


The provision of telework continues to require written agreement, either as an amendment to the original employment contract or as separate one. This agreement should notably rule the following topics:

  • Regime of permanence or alternation between distance and on-site work;
  • Place where the work is usually carried out and which will be considered, for all legal purposes, as the workplace;
  • Whether the equipment and systems are supplied by the company or acquired by the employee, as well as the person responsible for their installation and maintenance;
  • Frequency and form of face-to-face contacts;
  • Restrictions and conditions regarding the use of the equipment and systems for personal purposes (this provision is not mandatory, but it is advisable).

In addition, the possibility of refusing the provision of work through telework has been changed under the new legislation. The employee may refuse the employer telework proposal without having to justify such refusal. However, if the functions performed are compatible with the telework regime, the employee’s request can only be refused by the employer in writing and with a justification for such refusal.

Having this in mind and given the provisions of the new legislation, many companies are implementing remote work policies ruling and defining the activities and conditions under which telework may be accepted by the employer.

Duration and termination

According to the new law there are basically two models of telework agreement:

  • Fixed-term
    • Up to six months;
    • Renewable for same periods;
    • Either party may terminate the agreement in the first 30 days of execution;
    • Either party may terminate it by giving notice up to 15 days prior to the end of the term.
  • Undetermined duration
    • Either party may terminate the agreement in the first 30 days of execution;
    • Either party may terminate the agreement by giving notice 60 days prior the intended date of termination.

Working instruments and tools

Under the new legislation, there is now an express obligation for the employer to provide the employee with the equipment and systems necessary for teleworking and employee-employer interaction.

Moreover, the employer has also become responsible for bearing any additional expenses which the employee can prove to have incurred as a direct consequence of acquiring or using computer or telematic equipment and systems to carry out the work, including additional costs of energy and internet connection.

For the purposes of this law, additional expenses are (i) those arising from the acquisition of goods and/or services which the employee did not have before the agreement, as well as (ii) those determined by comparison with the expenses of the employee in the same month of the last year before the implementation of the telework agreement.

As one can easily identify, there are still remaining questions: Is the employee really going to be able to evidence increased costs given that many of them were already working from home during the pandemic lock downs? How should the payment of the increased costs be carried out when two or more cohabitants are teleworking? And how the privacy related rules will be taken into consideration if the employees are required to submit personal bills?

This will be an analysis on a case-by-case basis, considering the principles of non-duplication of compensation and of proportionality and adequacy. But nevertheless, ultimately one may claim that employees will be hardly able to evidence such increased costs. And if no such evidence is submitted by the employee, then no payment is due. As a matter of fact, as per the law such reimbursement will be due immediately after the cost being incurred by the employee, which is being already construed by some supervising authorities as the payment being due at the end of the month following the presentation of proof of expenses.

Given the difficulties for the employee in evidencing these cost, as well the potential increase in administrative tasks for the employer, many companies are implementing monthly remote work allowances. While these allowances will simplify the process, they will likely be considered as a taxable income for the employee, which otherwise will not happen with the reimbursement of expensed as provide in the law.

Employee’s Privacy

As was already the case under the previous legislation, the employer must respect the employee’s privacy, his working hours and his family’s rest times, as well as provide him with good working conditions, both from a physical and psychological point of view.

Although, when teleworking is carried out at the employee’s home, the employer can visit the employee’s workplace to control (i) work activity or (ii) work instruments. However, the employer is now obliged to respect a 24-hour prior notice for such visit and to obtain the employee’s agreement.

Furthermore, it is important to note that there are no longer any doubts (if there were yet any…) that capturing and using images, sound, writing, records or any other means of control that might affect the employee’s right to privacy are fully forbidden, being the breach of that rule considered a very serious offense.

Health and Safety at Work and Occupational Accidents

In the scope of the new legislation, a set of rules concerning safety and health at work, as well as occupational accidents was also foreseen:

  • Duty of the employer to promote occupational health examinations before the implementation of telework and, subsequently, annual examinations to assess the employee’s physical and mental aptitude to perform the work under this scheme.
  • Duty of the employee to provide access to the place where he/she performs work for the purposes of assessment and control of safety and health conditions at work, in a previously agreed period, between 9 a.m. and 7 p.m., within working hours.
  • For purposes of occupational accidents and diseases compensation, place of work is deemed to be the place chosen by the employee to habitually carry out his/her activity and working time is deemed to be all the time during which, demonstrably, the employee is providing his/her work to the employer.

Duty to abstain from contact

On 21 January 2021, the European Parliament approved a resolution stressing the importance of the right to disconnect and calling on the Commission to prepare a directive “that enables those who work digitally to disconnect outside their working hours”, being added as well that this directive “should also establish minimum requirements for remote working and clarify working conditions, hours and rest periods”. Being considered that “workers’ right to disconnect is vital to protecting their physical and mental health and well-being and to protecting them from psychological risks”, disconnecting is defined “as ‘not [engaging] in work-related activities or communications by means of digital tools, directly or indirectly, outside working time”. This resolution and the proposals therein raised concerns, but one thing was clear: the right to disconnect was in the centre of the discussions on the new work models.

In Portugal the new legislation also reflected this concern on the right to disconnect, but having a rather different approach and focus: the duty to abstain from contact.

According to the legislation in force as from January 1, 2022 the employer now becomes responsible to refrain from contacting the employee during his/her resting period, except in situations of force majeure. Furthermore, it was also foreseen that the breach of this rule, which applies to all employees (and not only to those teleworking) is considered a serious misdemeanour.

Since the provision for this duty was carried out in rather general terms, it is currently raising many practical questions in its implementation, including:

  • What should be considered as force majeure situation?
  • What should be understood as “employer” for this purpose?
  • What kind of contacts shall be included in the scope of this duty?

Cases of force majeure

The law does not provide for any definition of what should be included in this concept. However, given the general principles of the law, as well as some guidelines already issued for public administration, force majeure is an undetermined concept, which should be assessed on a case-by-case basis, covering, namely, situations in which the contact proves to be imperative, as it is indispensable to prevent or repair serious damages.

Concept of employer

Another issue which has also been raised, concerns the determination of the concept of employer for this purpose. Is the contact from a direct manager the only one considered for this law? And what if the contact is made by a colleague or even by someone of a different team?

Once again, the law does not provide any kind of definition, however it seems that not only all contacts from direct or indirect employee’s superiors should be considered, but also work colleagues under certain circumstances, namely when the contact of the colleague in question has the purpose of requesting information or collaboration in matters related to the provision of work.

Means of Contact

As regards the contacts covered, as a matter of principle all means of contact shall be included: e.g. telephone calls, text messages, e-mails, etc.

However, doubts remain as to whether or not sending, for example, an e-mail to an employee during his rest period but intending for that employee to reply or follow up on it only when he resumes work, constitutes a violation of the duty to abstain from contact.

This is yet a somehow controversial topic. There is already a public authority recently stating such conduct does not constitute a non-compliance with the duty to abstain from contact, provided that in the referred e-mail (i) no reply is requested, and (ii) no other immediate action is determined to the employee. Nevertheless, this approach, as said, is far from being generally accepted, despite the apparent strictness of the law not being often matching the flexibility required in today’s business life.

For this reason, the implementation of internal policies stressing notably that each employee is required only to answer during his/her work journey are more than advisable as tools to prevent breaches to the duty to abstain from contact.


While the first echoes on this new legislation were somehow disturbing, they were probably a bit exaggerated. No doubt that there is yet a long run to consolidate these work models and to add new worldwide trends such as increased flexibility or digital nomads. There are several questions pending to be answered and legal provisions to be clarified. Maybe there is a mentality twist that did not fully happen yet.  Nevertheless, this new legislation provides for a set of rules that ultimately will assist employers and employees with the implementation and organization of remote work. And in a way this legislation while needing to be perfected still claims for work life balance.


Modern technology is evolving at an exponential pace, based on Artificial Intelligence (hereafter, “AI”) and data-driven solutions. AI occupies a significant space in technological advances, several applications for machine learning and computer vision algorithms that continue to grow. While recognising AI and machine learning applications are the way forward into the future, it is also well-known that they require access to a large amount of data. Against this technological development, apparently, we can encounter data protection and privacy-friendly regulations. Not only the General Data Protection Regulation (GDPR), but we can identify several other data protection and privacy laws around the world that require companies to adapt to strict obligations. This raises the question of whether AI is compatible with such laws, GDPR being the standard. There is a clear tension between traditional data protection principles (e.g., purpose limitation, data minimization, and the limitation of automated decisions). A more flexible interpretation of the GDPR has been adopted for the development of AI and big data applications. However, this approach may be short-sighted in the long term and may hinder the development of such technologies. How to justify feeding massive volumes of data to form AI systems (high data consumption) without violating data protection legislation (e.g. principles to the fullest extent). Synthetic data has emerged as a privacy-friendly solution. We intend to launch and contribute to the debate about the advantages of synthetic data in the training of AI models, with a special focus on the interpretation of the GDPR.

What is Artificial Intelligence?

There are differences within the community when it comes to the definition of artificial intelligence systems (hereafter “AI”). Oversimplifying the concept, we can state that AI entails the ability or capacity of a machine to act purposefully, think rationally and deal effectively with its environment.[1] More in depth, AI can be defined in different subsets, such as symbolic learning, Machine Learning (hereafter “ML”) and reinforcement learning.

Symbolic learning is connected to explicit knowledge structures like ontologies, rule learning and therefore reasoning. This subfield was popular before the “recent” dawn of neural networks.[2]

Machine learning is characterised by statistical learning, benefiting from large volumes of data to learn models and is the present paradigm for AI.[3].

Lastly, we can also have reinforcement learning as a third subcategory of ML. The algorithm is defined by an optimal set of actions (a control policy) to achieve a certain goal, thus it is not told how to do a certain action but only receives positive or negative rewards that, in an iterative cycle, will bring the decision closer or further away from the primary goal.[4]

Full AI is thus characterized by a subset of operations that are typical of human intelligence, such as learning, adaptation, interaction, reasoning, problem-solving, knowledge representation, predicting and planning, autonomy, perception, movement and manipulation.[5]

The characteristics of this technology optimize human capability (reducing human shortcomings), having the potential to reduce task execution time, increase the productivity and efficiency of general actions, or the ability to increase the reading and analysis of mass data volumes (identifying and correlating the pattern).[6] Such capabilities are explained because AI is developed to discover correlations between data and build models, linking inputs to presumably correct responses (“predictions”). However, these predictions are only possible after the system is trained on vast sets of examples. AI is defined by a system that becomes “data-hungry” over time, and this requirement encourages data collection in a self-reinforcing spiral.[7]

AI data-driven technology

We can state that AI is thus data-driven technology that is hugely dependent on data. They require huge amounts of structured, semi-structured or non-structured sets of data (oversimplifying this matter, let’s call the need for “big data”).[8] The emergence of new technologies amounts to this reality, where mobile networks, social media and the internet of things generate even more data.[9]

The potential benefits are diverse and important for society.[10] However, as stated, the development of these applications is typically done via data hungry models, sometimes even when data for a specific purpose is not available. Solving data availability and abundance can thus be a challenge for the deployment of these technologies. Amassing the necessary data is not only technically demanding but can also entail several risks for privacy.[11]

The implementation of big data technologies can be defined into advanced data-driven strategies, based on the analysis and interpretation of verifiable and reliable data. Although data-driven management is a big challenge to entities, it enables the study and interpretation to get specific answers and present efficient solutions. Vast data sets are although difficult to manage using standard techniques, because of their special characteristics and definition on the several V’s: huge Volume, high Velocity, great Variety, low Veracity, and high Value.[12]  Understanding how big data technologies can be thus balanced in practice is not a simple task[13], especially when considering the emergence of new technologies in the context of IoT, social platforms, driverless cars, or smart cities. There are inherent risks for individuals of indiscriminate collection of data[14], analytics and decision-making based on AI techniques.

We can thus foresee that those data-driven technologies might contradict the inherent intention of data protection and privacy laws. How can we ensure purpose limitation data is collected indiscriminately and do not ensure data minimization or accuracy? It should also be considered that although such data can be created by individuals, most often are automatically collected by the system from the physical world or from computer-mediated activities.[15]

Current challenges: what about the data?

Recognising that AI can significatively improve human analytics and have a high and transversal impact on several sectors, it is also well-known that they require access to a large amount of data. While revealing clear economic and societal benefits[16], they entail challenging risks to fundamental rights and interests of individuals. We are thus confronted with a two-sided coin situation: where there are improvements or technological advances to be made, there are also a number of inherent risks that must be taken into consideration.[17]

Overreaching ethical and legal challenges arise and questions associated with ethics, fairness, transparency, accountability and explicability.[18] Aiming to tackle these challenges, the European Commission (hereafter “EC”) and the European Parliament (hereafter “EP”) repeatedly expressed the need for legislative action to ensure a well-functioning internal market for AI where benefits and risks are adequately addressed. Such an approach fundamentally aimed to ensure the development of secure, trustworthy and ethical artificial intelligence[19] and the protection of ethical principles.[20]

Seven key requirements for a trusted ecosystem for regulating AI have been proposed, in which “privacy and data governance” is highlighted.[21] It is understandable that if AI systems are to be designed around human rights, privacy and data protection should be taken as a priority.[22]

A proposal for a regulation laying down harmonized rules on AI[23] was adopted in the EU, establishing a balanced and proportionate horizontal regulatory approach while limited to the minimum necessary requirements to address the inherent risks. The regulation contains specific rules on the protection of individuals with regard to the processing of personal data, notably sets out the legal requirements for high-risk AI systems in relation to data and data governance, documentation and record keeping, transparency and provision of information to users, human oversight, robustness, accuracy and security. However, the regulation should be seen as complementary to existing data protection and privacy laws.

It is well known that building, training and testing AI models requires access to large and diverse data. While the proposal does not provide any direct legal solution to this question, such reality may conflict with data protection and privacy laws, notably the GDPR. It becomes clear that there is a tension between traditional data protection principles, for example purpose limitation, data minimization, and the limitation of automated decisions.[24] It is relevant to note that at the time of the entry into force of the GDPR AI systems were not a reality as they are today, and this piece of regulation mainly considered internet development.

It should be considered that data subjects should always be guaranteed full control over their own data, and that they are not allowed to process such data freely and without a legitimate basis of lawfulness. Imagine autonomous driving, where the algorithm collects data from the driver but also from sensors scanning the environment around (e.g. pedestrians or other vehicles). Can we consider that the data subjects have control over their data in these cases, while they are not even aware that their data is being processed?

While it is not our intention to conduct an in-depth analysis of all the issues surrounding this topic, the main idea that should be retained is that the development of such technologies is destined to raise data protection and privacy issues within the applicable EU legal framework. The European Data Protection Board already had the opportunity to emphasize the very high risk related to excessive data collection and the storage of such data over a long period of time, considering the development of new functionalities and, more specifically, those based on AI algorithms.[25]

Protecting the rights and interests of data subjects is therefore an essential part of developing IA systems. It is also recognized that the promotion of AI-driven innovation is closely linked to other data policy strategies, which establish trusted mechanisms and services for re-use, sharing and polling of data for the development of high-quality models.[26] That is why some ethical challenges inherent to AI are very similar to the ones raised by other technologies relying on big data, for example, on the abuse of data.[27] One might not realize that the collection and analysis of large datasets, often centralised, may lead to the re-identification of individuals, by means of linking datasets or inferring new data from existing datasets, nor that such collection is being used in another context than the one in which data was originally collected.

Such systems should therefore be underpinned by a logic of data ethics, outlining some principles that converge with data protection and privacy laws, for example, the purpose of the data processing is specifically identified, data is processed with respect for integrity, data sets are understandable, transparent and their use accountable, data should be open and individuals should have control over their own personal data.[28]

Synthesis as a possible solution

Synthetic data (hereafter, “SD”) is an artificially generated set of data, conceptually generated from a sample of real data, preserving its statistical properties without leveraging specific data records (the process can be then called as “synthesis”).[29] Data is generated through computer programs rather than real-world events. However, SD is not generated by any run-of-the-mill computer program. Rather, from algorithms that model the original statistical distributions and structure of real datasets. Even though the output data might be considered “fake”, it still maintains utility for training models.[30]

Synthetic datasets can be used to fit specific needs, such as testing new tools or just to share data. Synthesis is thus an emerging privacy-enhancing technology that can enable inexpensive access to unlimited new samples of realistic data based on a model that is trained on real data.[31] More recently advances in AI have enabled the creation of even more realistic synthetic datasets, with deep learning models leading the innovation in this field. These models can be automatically trained on available data, and then be used to generate new, unlimited SD samples.

Synthetic generators, such as the ones based in the celebrated Generative Adversarial Network architectures,[32] are a flexible and easy to use AI based solution that enables organizations to generate highly realistic synthetic events. SD can thus play a key role in the adoption of new technologies, especially considering that privacy and data protection laws have made it even more difficult to access and use real data for training AI models.

Defining types of synthetic data

Conceptually SD is not real data, but data that is generated from real data and maintains the same statistical properties.[33] There are three types of SD.[34] Those that are generated from real/actual data sets, those that do not use real data, and hybrids of the first two types.

Data synthesis based on real data is the most commonly used since the applied models are often domain agnostic, its development is less domain knowledge intensive, they directly adapt to the context from where the data is collected, often regardless of complexity.

Best in class: synthesis as technical measure

In contrast to other privacy-enhancing technologies that, to some extent, render information to be non-personal (de-identified data)[35] SD is not real data related to real individuals and there is no link between records in a SD set and records in the original real data set.

Recital 26 of the GDPR states that the regulation does not apply to “information which does not relate to an identified or identifiable natural person or to personal data rendered anonymous in such a manner that the data subject is not or no longer identifiable”.[36] Data synthesis would then benefit from the same reasoning as the case for not needing to comply with further obligations for anonymization under GDPR.[37] However, depending on the synthesis process applied, the need to ensure initial data creation and testing must comply with applicable laws. Notably, the process of anonymization is itself a data processing activity that needs to comply with data protection laws.[38]

Anonymization is an important measure to ensure that data subjects are not re-identified, limiting the risks over their rights and freedoms while enabling us to make the information available to the public. Although, successful anonymization may be practically impossible for complex dataset.[39]

We should also consider risk-based de-identification, as the process used to prevent personal identifiers from being connected with information.[40] In some cases, de-identification may be a solution, in particular for AI training models, notably because training data need not include directly identifying data.[41]

Considering the above, the identified advantage is that SD uses characteristics of the real data (or not) to generate new “fake” data. While entailing statistical models of distributions and structure of the real datasets, SD are similar to the original one in terms of granularity. Nevertheless, SD should not be held as personal data, since the information is not linked to actual data subjects.

Tackling current challenges

Among the many challenges that were identified, the dependency on data is one of the greatest.

Applying synthesis as a privacy measure can help tackle some issues when it comes to data and data protection laws, mainly the compliance with fundamental data protection principles.[42] This solution could avoid the need to promote a flexible interpretation of the GDPR so as not to hamper the use of personal data for AI purposes.

Repurposing might not be compatible with new purposes that are different from those for which the data were originally collected.[43] This constant analysis and balancing of rights and interests may be difficult or even impossible with the development of AI, even more so when we consider the learning capabilities of such systems.

There is also a tension between the idea of data minimisation, according to which personal data should be adequate, relevant and limited to what is necessary in relation to the purposes for which they are processed.[44] It is relevant to consider that the principle of data minimization implies that entities should identify the minimum amount of personal data to fulfill their purposes, and no more.[45] There is thus a clear tension with the concept of big data and data analytics and this principle. Considerations of ways in which this tension can be reduced are bound to involve long-term practical issues.[46] Constant assessments of proportionality in data inclusion, limiting processing to statistics, and avoiding making decisions about individuals can become too much of a burden when ensuring the evolution of training models and the development of the system. In close connection with the data minimisation principle is the lawful ground for processing, notably AI prediction is provided as a service.[47] It should be considered that using personal data for performing or entering a particular contract does not cover the subsequent use of such data for purposes of business analytics, or possible inputs to a predictive-decisional model concerning the data subject.[48] Data minimisation should not, however, be confused with anonymization. Although there are conceptual and technical similarities between the concepts, privacy-preserving techniques normally means that certain data used in AI are rendered pseudonymous or anonymous.[49]

The risk-based approach of data protection laws entities to assess their risk appetite for AI development, for example, considering the implementation of technical measures in the context of particular circumstances.[50] This obligation can in some cases hinder the evolution of such systems, forcing governments to create special regimes like “sand-boxes” for high-risk technologies. The expansion of these regulatory frameworks emerges to tackle current challenge of this “new” technologies. Inserted in this environment, in a regulatory tension, companies may enjoy a wide interaction with regulators for the controlled development of services and products offered. This way, at a later and safer moment, regulatory agencies will be able to choose “if”, “how” and “when” to grant definitive authorizations for the entry and real performance of these companies in the consumer market.

The Portuguese Government established in a Council of Minister Resolution No. 29/2020, the bases for the establishment of the so-called “Technological Free Zones” (ZLT) “for the testing and experimentation in real environment in the country of any new technologies and solutions”. In generic terms, this space is intended to be a general and intersectoral structure for the experimentation of innovative technologies. In other terms, this would be a representation corresponding to the concept of the regulatory sandbox regime. The legal framework for the ZLTs, was later approved by DL 67/2021, of 30 July, in which the regime and governance model for promoting technology-based innovation through the creation of these zones was established.

Compliance with applicable laws should not be sacrificed to promote the creation of training sets and the construction of algorithmic models in which the resulting AI systems could pose a risk to the rights and interests of data subjects. Since privacy regulations may not apply to SD, violation of the legislation or even the threat of a data breach would not be an issue, at least for the associated use cases.[51]

SD may enable the use of datasets not restricted by legal and privacy concerns, allowing the development of new innovative solutions without risking data subjects re-identification. The opportunity entails that organizations can become compliant with data protection and privacy laws while being innovative.[52]

Leveraging SD can also bost data economy. Organizations are able to share their data without incurring the risk of exposing personal data.[53] The use of it accelerates and eases the data sharing process between entities, helping the creation of a trustful global data economy.[54] Sharing data may empower crowd intelligence, leading to new research and innovation.[55]

Developing and testing software solutions also benefit from this solution. SD eases the creation of high volumes of realistic data for other environments such as development or even testing environments.[56] SD generators are by design conceived to be privacy preserving, these allow sampling for rich datasets boost the development and accuracy of the software development, mitigating the risk of future bugs, improving the testing process and an integration flow less prone to errors.[57]

SD can also support AI research and new methods to predict a certain disease. SD generated from original and real patients’ data, can solve some privacy concerns while keeping the datasets characteristics and features that allow us to identify a certain disease or support a certain diagnosis.[58]


We can conclude that the use of SD may allow entities to occlude personal data from generated datasets with less impact on their utility and more effectiveness when compared other technical measures. Classical anonymization techniques are often applied to real data but are limited in blocking re-identification events, are laborious to implement, and also require some degree of expertise to identify. Differential privacy techniques may allow avoiding these pitfalls and can be directly integrated in synthesizer training pipelines leading to synthesizer models with a controlled privacy guarantee and the least impact in utility.

This unique feature can potentially lay grounds for compliance with applicable law in data protection and privacy, notably the GDPR. In some cases, it would not be necessary to promote a flexible interpretation of the law so as not to undermine the use of personal data for AI purposes.

Beside the positive outlook in what comes to privacy protection, SD can open up AI models to the possibility of approximating real distributions in a way that would not otherwise be possible, or economically feasible when relying solely on limited real world samples. Data generators can interpolate from real observations, in some cases even extrapolation capacity can be observed. The usual constraints of collection and labeling of real-world data are not hindrances since unlimited samples can be obtained from the modeled distributions without incurring additional costs. These tools can also be used to remove unbalance from available real-world samples, which can be used to remove bias, ensuring that production models respond correctly to rare catastrophic events and other important challenges behind data applications.

The ability to produce virtually infinite synthetic samples synergizes with privacy in a best of both worlds scenario – higher privacy to utility tradeoffs penalizes utility, however, with unlimited potential of creating new data points, this drawback can effectively be mitigated in AI application development.

In the context of this paper, there are several interesting doors that SD can open. However, entities should be aware that this technology is subject to possible limitations like any other emerging technology and should not privilege its exclusive use over others technological privacy-friendly measures.

[1] Rembrandt Devillé, Nico Sergeyssels and Catherine Middag, “Chapter 1, Basic Concepts of AI, for Legal Scholars”, Artificial Intelligence and the Law, Jan De Bruyne and Cedric Vanleenhove (eds.), Intersentia, p.2.

[2] Minsky, M. L.. “Logical versus analogical or symbolic versus connectionist or neat versus scruffy”, AI magazine, 1991, 12(2), p.34; Blazek, P., How Aristotle is Fixing Deep Learning’s Flaws, 2022; and Marcus, G., “The next decade in AI: four steps towards robust artificial intelligence”, 2020.

[3] Dinsmore, J., “The symbolic and connectionist paradigms: closing the gap. Psychology Press”, 2014.

[4] Rembrandt Devillé, Nico Sergeyssels and Catherine Middag, “Chapter 1, Basic Concepts of AI, for Legal Scholars”, Artificial Intelligence and the Law, Jan De Bruyne and Cedric Vanleenhove (eds.), Intersentia, p.6-7.

[5] Agência para a Modernização Administração (ama), “GuIA, Guia para uma inteligência artificial ética, transparente e responsável na administração pública”, p. 10.

[6] For some successful applications of ML see Wang, W., Ye, Z., Gao, H., & Ouyang, D., “Computational pharmaceutics-A new paradigm of drug delivery”, Journal of Controlled Release, 2021, 338, p.119-136

[7]  European Parliament Research Service (EPRS), “The impact of the General Data Protection Regulation (GDPR) on artificial intelligence”, Scientific Foresight Unit (STOA), PE 641.530 – June 2020, Section I.

[8] Sarker, I.H., “Machine Learning: Algorithms, Real-World Applications and Research Directions”, SN COMPUT. SCI. 2, 2021, p. 160.

[9] Bernard Marr, “How Much Data Do We Create Every Day? The Mind-Blowing Stats Everyone Should Read”, Forbes, 2018.

[10] Information Commissioner’s Office, “Big data, artificial intelligence, machine learning and data protection”, Data Protection Act and General Data Protection Regulation, 20170904, Version:2.2, pp. 15 to 18.

[11] Ericsson, “Privacy in mobile networks – How to embrace privacy by design”.

[12] European Parliament Research Service (EPRS) (…) p.4.

See also, Agência para a Modernização Administração (ama) (…) p. 26.

[13] Vestoso, Margherita. 2018. “The GDPR beyond Privacy: Data-Driven Challenges for Social Scientists, Legislators and Policy-Makers” Future Internet 10, no. 7: 62.

[14] European Parliament Research Service (EPRS) (…) p.18.

[15] Ibidem, p.4.

[16] European Commission, “White Paper: on artificial Intelligence – A European approach to excellence and trust”, COM(2020) 65 final, Brussels, 18.2.2020, p.2.

[17] Concerning this idea see European Commission, “White Paper On Artificial Intelligence -A European approach to excellence and trust”, Brussels, 19.02.2020, COM(2020) 65 final.

[18] Chapter 3 Setting the scene: on AI ethics and regulations, Artificial Intelligence and the Law, Jan De Bruyne and Cedric Vanleenhove (eds.), Intersentia, p.49.

[19] European Council, Special meeting of the European Council (1 and 2 October 2020) – Conclusions, EUCO 13/20, 2020, p. 6.

[20] As a general reference see the footnote references Concerning this idea see European Commission, “White Paper On Artificial Intelligence -A European approach to excellence and trust”, Brussels, 19.02.2020, COM(2020) 65 final.

[21] The High-level Expert Group on Artificial Intelligence defined seven key requirements for a trustworthy AI: (i) human agency and oversight; (ii) technical robustness and safety; (iii) privacy and data governance, (iv) transparency; (v) diversity, non-discrimination and fairness; (vi) societal and environmental wellbeing; and (vii) accountability (cf. High-Level Expert Group on Artificial Intelligence, set up by the European Commission, “Ethics Guidelines for Trustworthy AI, 201). See also European Parliament Research Service (EPRS) (…) p.31. And also the Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, “Building Trust in Human-Centric Artificial Intelligence”, Brussels, 8.4.2019, COM(2019) 168 final.

[22] Trustworthy AI systems should thus entail the following five dimensions: Accountability, ensuring that the systems are accountable, secure and auditable. Transparency, ensuring the visualization of their components and the procedures applied. Explicability, ensuring that the systems can be understood by the explanation provided. Fairness, ensuring that the systems are fair and non-discriminatory. Ethics, ensuring that the system offers mitigations to deal with ethical bias (Cf. Agência para a Modernização Administração (ama) (…) p. 33.)

[23] Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence ACT) and amending certain union legislative acts {SEC(2021) 167 final} – {SWD(2021) 84 final} – {SWD(2021) 85 final}, COM(2021)206 final, 2021/0106 (COD), p.3.

[24] Zarsky, Tal Z.. “Incompatible: The GDPR in the Age of Big Data”, Seton Hall Law Review, 2017. Vol. 47:995, pp. 1003 – 1004.

[25] European Data Protection Board (EDPB), “Guidelines 01/2020 on processing personal data in the context of connected vehicles and mobility related applications”, version 2.0, adopted on 9 March 2021.

[26] Proposal for a Regulation (…) p.5.

[27] For a deeper understanding of these ethical challenges, especially on data abuse, is recommended to consult Michiel Fierens, Stephanie Rossello and Ellen Wauters, “Chapter 3 Setting the scene: on AI ethics and regulations”, Artificial Intelligence and the Law, Jan De Bruyne and Cedric Vanleenhove (eds.), Intersentia, p.51.

[28] Agência para a Modernização Administração (ama) (…) pp.. 22-23.

[29] Khaled El Eman, Lucy Mosquera & Richard Hoptroff, “Practical Synthetic Data Generation, Balancing Privacy and the Broad Availability of Data” (2020), p.1.

[30]  Li, H., Xiong, L., & Jiang, X. (2014). Differentially Private Synthesization of Multi-Dimensional Data using Copula Functions. Advances in database technology : proceedings. International Conference on Extending Database Technology, 2014, 475–486.

[31] Although, the generation of synthetic data is something that is already known in the market for the last decades, mainly through techniques such as modeling a multivariate probability distribution for a given data set and then sampling new data.

[32] A. Creswell, T. White, V. Dumoulin, K. Arulkumaran, B. Sengupta and A. A. Bharath, “Generative Adversarial Networks: An Overview,” in IEEE Signal Processing Magazine, vol. 35, no. 1, pp. 53-65, Jan. 2018 (doi: 10.1109/MSP.2017.2765202).

[33] Khaled El Eman, Lucy Mosquera & Richard Hoptroff, “Practical Synthetic Data Generation, Balancing Privacy and the Broad Availability of Data”, 2020, p. 1.

[34]  Ibidem.

[35] An example is the removal of private health information from medical records (Uzuner, Ö., Luo, Y., & Szolovits, P., “Evaluating the state-of-the-art in automatic de-identification”. Journal of the American Medical Informatics Association, 14(5), 2007, pp. 550-563)

[36] Randy Koch, GDPR, CCPA and beyond: How synthetic data can reduce the scope of stringent regulations, 2020.

[37] El Emam, “Accelerating AI with synthetic data”.

[38] Barta, G. (2018). Challenges in the compliance with the General Data Protection Regulation: anonymization of personally identifiable information and related information security concerns. Knowledge–economy–society: business, finance and technology as protection and support for society, chapter 11.

[39] Rocher, L., Hendrickx, J. M., & De Montjoye, Y. A., “Estimating the success of re-identifications in incomplete datasets using generative models”, Nature communications, 10(1), 2019, p.1-9

[40] El Emam, K. (2010). Risk-based de-identification of health data. IEEE Security & Privacy, 8(3), p.64-67

[41] For a discussion on this matter, and mitigation strategies, see Google’s Considerations for Sensitive Data within Machine Learning Datasets.

[42]  Charline Daelman, “Chapter 6, AI through a Human Right Lens. The Role of Human Rights in Fulfilling AIs Potential”, Artificial Intelligence and the Law, Jan De Bruyne and Cedric Vanleenhove (eds.), Intersentia, p.123.

[43] European Parliament Research Service (EPRS) (…) pp.45 to 47. Also, see WP29, Opinion 03/2013 on purpose limitation, adopted on 2 April 20213, 00569/13/EN, WP 203.

[44] Ibidem, pp.47 to 48.

[45] Information Commissioner’s Office (ICO), Guide to the General Data Protection Regulation (GDPR), 2021, pp. 28-31.

[46] European Parliament Research Service (EPRS) (…) pp.47 to 48.

[47] Information Commissioner’s Office (ICO), Guidance on the AI auditing framework, Draft guidance for consultation, 2019, p. 23.

[48] Ibidem p. 50

[49] Ibidem p. 83.

[50] Ibidem p. 13.

[51] Randy Koch, GDPR, CCPA and beyond: How synthetic data can reduce the scope of stringent regulations, 2020.

[52] Anderson, J. W., Kennedy, K. E., Ngo, L. B., Luckow, A., & Apon, A. W. (2014, October). Synthetic data generation for the internet of things. In 2014 IEEE International Conference on Big Data (Big Data) (pp. 171-176). IEEE

[53] El Emam, K., & Hoptroff, R., “The synthetic data paradigm for using and sharing data”, Cutter Executive Update, 19(6), 2019.

[54] Steinhoff, J. Towards a Political Economy of Synthetic Data: The Possibility of a Data-intensive Capitalism That is Not a Surveillance Capitalism

[55] Piwowar, H. A., Becich, M. J., Bilofsky, H., Crowley, R. S., & caBIG, “Data Sharing and Intellectual Capital Workspace”, Towards a data sharing culture: recommendations for leadership from academic health centers. PLoS medicine, 5(9), 2008, p. 183.

[56] Kamaev, A. N., Smagin, S. I., Sukhenko, V. A., & Karmanov, D. A., “Synthetic data for AUV technical vision systems testing”, In CEUR Workshop Proceedings Vol. 1839, 2017, p. 126-140.

[57] Whiting, M. A., Haack, J., & Varley, C., “Creating realistic, scenario-based synthetic data for test and evaluation of information analytics software”, In Proceedings of the 2008 Workshop on beyond time and errors: novel evaluation methods for Information Visualization, 2008, April, p. 1-9.

[58] This dataset is composed of 3 different types of input features: (i) objective, based in factual information; (ii) examination, translating the results of medical examination; and subjective, that is the information given by the patient.