-
What are your countries legal definitions of “artificial intelligence”?
As of today, there is no legal definition for “artificial intelligence” (AI) under French legislation. The main definition that may be relied upon stem from the draft AI Act: “a system that is designed to operate with elements of autonomy and that, based on machine and/or human-provided data and inputs, infers how to achieve a given set of objectives using machine learning and/or logic – and knowledge based approaches, and produces system-generated outputs such as content (generative AI systems), predictions, recommendations or decisions, influencing the environments with which the AI system interacts”. OECD defines AI as “an automated system that, for a given set of objectives defined by humans, is capable of making predictions, formulating recommendations, or making decisions that impact real or virtual environments. AI systems are designed to operate at various levels of autonomy.”. In France, the data protection authority (CNIL) issued guidelines which define artificial intelligence as “a logical and automated process generally based on an algorithm and capable of performing well-defined tasks.”
-
Has your country developed a national strategy for artificial intelligence?
France’s strategy for AI is outlined within the framework of the France 2030 Program, which identifies AI as a priority for the country’s future development This strategy incorporates the recommendations from a 2018 “Villani Report”, shaping France’s approach to AI and guiding specific actions and initiatives to foster innovation, ensure responsible data usage, nurture AI talent, and promote ethical practices.
As part of this program, French President announced a more specific plan at Vivatech in June 2023 to support the AI field. The “AI-cluster” initiative, endowed with €500 million, is meant to strengthen national centers of excellence in education and research and increase the number of AI specialists in France by 2030.
Funding will also be allocated to enhance the Jean Zay supercomputer for scientific research and co-finance the development of a new exascale-class supercomputer by 2024-2025, in collaboration with the European Union.
Lastly, an open call for projects, with a budget of €40 million, will establish digital commons for generative AI, including datasets that reflect French and European values such as privacy protection, freedom of expression, and use of French language. This is meant to mitigate potential biases induced by the use of foreign datasets.
-
Has your country implemented rules or guidelines (including voluntary standards and ethical principles) on artificial intelligence? If so, please provide a brief overview of said rules or guidelines. If no rules on artificial intelligence are in force in your jurisdiction, please (i) provide a short overview of the existing laws that potentially could be applied to artificial intelligence and the use of artificial intelligence, (ii) briefly outline the main difficulties in interpreting such existing laws to suit the peculiarities of artificial intelligence, and (iii) summarize any draft laws, or legislative initiatives, on artificial intelligence.
Existing laws & rules. Currently, the only effective specific rule pertains to the criminal liability of autonomous car manufacturers in the event of a road accident, as established by Order (second level legislation) No. 2021-443 of April 14, 2021. This regulation outlines the liabilities of the AI system and the human operator in autonomous mobility services. In the absence of other specific legislation, existing common law principles such as liability for defective products, liability for things, vicarious liability, and liability for animals, may be applied to cases involving AI systems. However, the unique characteristics of AI, such as autonomy and complexity, pose challenges for traditional liability rules.
Draft regulation initiatives on AI. Efforts to regulate AI are primarily focused on European level to promote a unified approach to governing artificial intelligence and, when relevant, harmonize regulations. The draft AI Act is a regulatory framework proposed by the European Union currently in the final negotiation phase following the European Parliament’s adoption of its position in June 2023. It will establish rules and standards for AI systems in specific sectors. It addresses concerns related to fundamental rights and protection of individuals from discriminatory or harmful AI practices. In addition, a directive proposal dated 28 September 2022, expected to be further discussed after the adoption of the AI Act, aims to adapt liability regimes to AI systems and clarify the burden of proof in cases involving AI.
Guidelines. Pending comprehensive legal regulations, guidelines have been developed at both a European and national levels, to establish ethical standards and principles.
In 2019, the European Union’s High-Level Expert Group on AI, published a report titled “Ethics Guidelines for Trustworthy AI.”, putting forward ethical principles and emphasizing transparency, accountability, and the robustness of AI systems.
Complementing this, the CNIL issued further guidance through various documents and publications that should be adhered to when deploying AI systems.
Lastly, sectoral authorities, such as the banking supervisor (ACPR) and the Ethics Committee of the Ministerial Delegation for Digital Health in France, based on recommendations from the European Commission, WHO, the OECD, and UNESCO, also provided guidance to address the impacts of AI within their respective fields.
-
Which rules apply to defective artificial intelligence systems, i.e. artificial intelligence systems that do not provide the safety that the public at large is entitled to expect?
France implemented a special liability regime for defective products based on the EU Directive 85/374/EEC of 25 July 1985. This regime applies to movable goods (including those incorporated into other goods or buildings) that fail to meet expected safety standards, regardless of whether there is a contractual relationship between the victim and the producer. AI systems as such are not expressly covered by these provisions, but can in principle, when incorporated into a product, cause damage.
However, this framework may not fully address the complexities of AI-related defects. For instance, the exemption of liability for development risks may not be relevant in the case of AI, and the determination of the liable party can be challenging given that the current focus is primarily on the producer. To address this, the European Commission proposed a revision of Directive 85/374/EEC to explicitly include AI systems and AI-enabled goods as “products” within its scope. The proposal expands liability to all “economic operators” in the supply chain (including manufacturers/producers, the provider of a related service, the authorized representative or distributor, and even online platforms), addresses challenges in determining the responsible party, and introduces compensation for data loss.
Overall, the existing legal arsenal provides some mechanisms to address the consequences of a faulty AI but further refinements and adaptations remain necessary.
Liability concepts such as, “fault-based liability” and “liability for things” can also apply.
The General Data Protection Regulation (GDPR) can also be invoked in case of data breaches and privacy violations related to AI.
-
Please describe any civil and criminal liability rules that may apply in case of damages caused by artificial intelligence systems.
Regarding bodily or material harm caused by a faulty AI, liability concepts such as the special liability for defective products, “fault-based liability” and “liability for things” can apply.
In case of personal data leak, GDPR can be invoked, awarding a right for data subjects to seek compensation for damages.
-
Who is responsible for any harm caused by an AI system? And how is the liability allocated between the developer, the user and the victim?
Under the classic extra-contractual liability regime, the person responsible for damage is determined based on the fault that caused the harm. This fault can be attributed to the supplier, user, or any other economic operator in the chain. And if multiple parties contribute to the damage, they are held individually liable for their respective share. In cases where determining individual responsibility is impossible, joint and several liability may be imposed. The victim’s own fault or events such as force majeure, may partially or completely absolve the party responsible for the damage.
Under the regime for defective products of Directive 85/374/EEC, as transposed in France, if an AI system causes damage due to a design defect, the producer or developer may be held accountable. If the product is non-European and imported, the importer within the EU is also considered a producer and subject to liability. If none of the producer, importer, or developer can be identified, the seller may be held responsible.
Under the specific autonomous cars regime, the car manufacturer bears criminal liability under Order n°2021-443 dated 14 April 2021 if an accident occurs while the vehicle is in automatic mode.
In other no-fault liability regimes, such as those mentioned earlier, the responsible parties are typically the owners or custodians of the entity involved, whether it’s a thing, person, or animal.
-
What burden of proof will have to be satisfied for the victim of the damage to obtain compensation?
Typically, civil liability requires the fulfillment of three conditions: fault, damage, and a causal relationship between the fault and the damage. The burden of proof usually rests on the victim of the damage.
In cases involving defective products, fault is replaced by the presence of a product defect, which arises when the product fails to meet expected safety standards. It is the victim’s responsibility to prove the defectiveness of the product and establish the causal connection to the damage. No presumption of liability exists within this framework.
In a no-fault liability regime, the victim must demonstrate that an incident resulted from a thing, person, or animal, and establish the causal link between that incident and the damage. Furthermore, to hold the responsible party liable, a relationship between the guardian and the thing, person, or animal must be established. Presumptions have been created by law and case law to facilitate the determination of liability, but none apply to AI-related damages yet.
-
Is the use of artificial intelligence insured and/or insurable in your jurisdiction?
Currently, there is no specific insurance coverage tailored for the use of AI. However, traditional insurance policies such as professional indemnity, civil liability, cybersecurity, product liability, and directors’ and officers’ liability can be customized to address the risks associated with AI systems by providing extended coverage.
-
Can artificial intelligence be named an inventor in a patent application filed in your jurisdiction?
An AI cannot be named an inventor in European Patent applications, as confirmed by the Board of Appeal of the European Patent Office on 21 December 2021 (J8/20).
The French Intellectual Property Office (INPI) has not yet ruled on whether an artificial intelligence can be named an inventor in French patent applications. However, this does not seem compatible with INPI guidelines which state that the inventor is a “natural person” and Article R. 612-10 of the French Intellectual Property Code referring to the “last name, first name and domicile of the inventor”.
-
Do images generated by and/or with artificial intelligence benefit from copyright protection in your jurisdiction? If so, who is the authorship attributed to?
French courts have not yet ruled on whether images generated by AI benefit from copyright protection, but this seems unlikely as only original creations can be copyright-protected. First, the creation criterion seems to require human intervention: the French Supreme Court ruled that legal entities cannot be authors, implying that only natural persons can (Cass. civ. 1, 15 January 2015, 13-23.566). Second, the originality criterion requires the work to reflect the imprint of the author’s personality, which would exclude creations generated by machines.
In contrast, images generated with the assistance of artificial intelligence may arguably benefit from copyright protection if there is a genuine human creation that goes beyond mere instructions to an AI (e.g. the image generated by the AI is reworked or integrated into another work). If the final image is original, it would be eligible to copyright and it is likely that the author(s) would be the person(s) who reworked it.
-
What are the main issues to consider when using artificial intelligence systems in the workplace?
When implementing AI systems in the workplace, there are several important considerations to address:
- Accuracy and Reliability: Employers must ensure that AI systems are accurate and reliable. Regular monitoring and evaluation should be conducted to identify and rectify biases, errors, or limitations that could impact system outcomes. Adequate training and support should be provided to employees regarding AI capabilities, limitations, and potential risks. Employees should be aware of their rights and responsibilities when interacting with AI systems.
- Decision-Making and Outcome Responsibility: As AI systems play a larger role in decision-making, it is crucial to establish guidelines and procedures for accountability. Human oversight should be present, and mechanisms should be in place for intervention or review of decisions made by AI systems.
- Data Security and Confidentiality: AI systems often rely on sensitive employee data, requiring robust measures to protect it against breaches or unauthorized access. Compliance with data protection regulations is then essential.
While such cases have not been reported in France, other jurisdictions have experienced incidents related to employees’ use of AI. For example, Samsung took action in South Korea after a data leak caused by employees using generative AI. To prevent further breaches, Samsung implemented restrictions on downloading capabilities and initiated an investigation. The company is also exploring the development of its own AI systems to enhance data security.
-
What privacy issues arise from the use of artificial intelligence?
Data Privacy: The main privacy concerns surrounding AI is the potential for data breaches and unauthorized access to personal data considering AI systems often require vast amounts of data to train and operate effectively. This also challenges individuals’ ability to provide informed consent.
Surveillance and Tracking: AI technologies, such as facial recognition and predictive analytics, can be used for extensive surveillance and tracking of individuals, leading to concerns about invasion of privacy.
Bias and Discrimination: AI systems may perpetuate existing biases and discrimination they could inherit from training data, leading to unfair or discriminatory outcomes. This can disproportionately impact certain groups and perpetuate existing societal biases.
-
What are the rules applicable to the use of personal data to train artificial intelligence systems?
The GDPR serves as the primary legislation for personal data protection in France and Europe, encompassing AI systems through its technology-neutral approach. The creation of training databases containing personal data must adhere to the GDPR’s requirements and principles. Compliance with personal data protection laws from the inception of an AI system is mandated by the GDPR’s privacy by design principle. This includes ensuring lawful, purposeful, minimal, and secure processing of personal data during AI system training.
Furthermore, data processing during the training phase should enable individuals to exercise their rights, such as access, rectification, erasure, data portability, and the right to object to automated decision-making.
-
Have the privacy authorities of your jurisdiction issued guidelines on artificial intelligence?
Building upon the 2019 EU’s High-Level Expert Group on AI’s “Ethics Guidelines for Trustworthy AI.”, the CNIL issued numerous contents and guidance on AI through various documents and publications that should be adhered to when deploying AI systems.
The CNIL’s guidelines highlight the importance of ensuring transparency, fairness, non-discrimination and accountability in AI applications and advocate for organizations to conduct impact assessments to identify and mitigate potential risks, particularly regarding data protection and privacy.
They also emphasize the need for individuals’ informed consent when processing personal data using AI systems. Organizations are encouraged to implement measures that guarantee individuals’ rights, such as access to and rectification of their personal information.
Lastly the guidelines stress the importance of avoiding decisions solely based on automated processing. Data minimization and security are underscored, advising organizations to collect necessary data only and implement robust security measures.
In its 2023’s action plan, the CNIL demonstrates its intention to regulate key sectors in AI, with a focus on augmented cameras, generative AI, large language models, and related applications like chatbots, and tackles four key pillars: understanding the functioning of AI systems and their impacts on individuals, enabling and regulating the development of privacy-respecting AI, fostering and supporting innovative actors in the AI ecosystem in France and Europe, and auditing, controlling, and safeguarding individuals in relation to AI systems. These efforts will also contribute to preparing for the implementation of the ongoing discussions around the European AI Act project.
-
Have the privacy authorities of your jurisdiction discussed cases involving artificial intelligence?
In a decision of 17 October 2022, the CNIL imposed a fine of 20 million euros and ordered Clearview AI not to collect and process data on individuals located in France without any legal basis, and to delete the data of these individuals, after responding to requests for access it received. Clearview AI is a company that has developed facial recognition software using publicly available photographs and videos from the Internet. They offer law enforcement authorities access to their extensive image database through a search engine, enabling them to identify individuals based on photographs using facial recognition technology.
-
Have your national courts already managed cases involving artificial intelligence?
To date, French courts have not had the occasion to deal with cases involving artificial intelligence systems. However, a significant case involving Google and its autonomous suggestion system, Google Suggest, was brought before the French Supreme Court. In 2011, a company sued Google when the term “swindler” appeared in the search suggestions associated with their name. Initially, Google was found guilty of “public insults” as the judges considered that Google was not totally neutral in its data processing and could not hide behind the automatic nature of the process, as there was a “possibility of human control over the functionality“. The Supreme Court overturned the decision and ruled that Google could not be held liable for the automatic and random process of its suggestion functionality, as it did not have the intention to create or endorse the suggestions. This decision represents a reversal of previous case law, where Google was held responsible for the content in its suggestions based on pre-sorting and the potential for subsequent control.
-
Does your country have a regulator or authority responsible for supervising the use and development of artificial intelligence?
In the absence of specific legislation, there is currently no dedicated authority solely responsible for addressing issues related to AI. However, the CNIL assumes a cross-cutting position and acts as the leading authority in this field. The designation of the CNIL as the competent authority for overseeing the implementation of the AI Act is recommended by senators in a European resolution proposal in March 2023. Other sector-specific authorities, such as the ACPR, in the bank and insurance sector, and the Haute Autorité de Santé, in the health sector, also provide guidance or literature on the subject.
-
How would you define the use of artificial intelligence by businesses in your jurisdiction? Is it widespread or limited?
According to a study conducted in May 2023 by BVA, a consulting firm, AI is widely adopted across various industries in France. Over 35% of companies with ten or more employees are either using AI or are in the process of implementing it. However, the adoption rate varies across sectors.
The agriculture sector leads in AI usage, with 58% of companies employing AI technologies, with the industry sector following closely at 50%. Finance and commerce sectors also demonstrate significant adoption rates of 44% and 40% respectively. On the other hand, AI use remains relatively low in the construction and personal services sectors, with both sectors reporting adoption rates below 30%.
Larger companies with at least 200 employees tend to have higher AI adoption rates, at 45%.
In terms of applications, decision support systems, natural language processing, and robotics are the primary areas where AI is used in France.
-
Is artificial intelligence being used in the legal sector, by lawyers and/or in-house counsels? If so, how?
The legal sector in France is not one in which artificial intelligence is widely used. Nevertheless, its development is underway, and there are several uses for it, in consulting, litigation and administrative tasks.
- Legal Research: Tools like Westlaw Edge or Doctrine use AI algorithms to analyse vast legal databases and provide lawyers with case law, statutes, and legal opinions for their research.
- Contract Analysis: platforms such as Della and Seal Software use AI to extract key information from contracts, flag potential risks, and provide contract summaries.
- Predictive Analytics: Companies like Case Law Analytics, Lex Machina and Predictice use AI to analyze legal data, including court records and case outcomes, to generate predictive analytics and help lawyers assess the likelihood of success in litigation or settlement negotiations.
- Contract Management: Solutions like OpenLaw provide AI-powered contract management platforms. Their solution streamlines the contract creation and review process, enabling collaboration, version control, and automated contract generation based on predefined templates.
- Legal Chatbots: Various chatbots are used in France, such as DemanderJustice, leveraging AI technology to provide online legal advice, address common legal queries, and offer information on consumer rights, tenancy disputes, family law, and more.
- E-Discovery: AI-powered e-discovery platforms such as Relativity employ machine learning algorithms to process and analyze large volumes of electronic documents, identifying relevant documents based on context, keywords, and patterns.
In early 2023, certain law firms announced a partnership with Harvey, an artificial intelligence platform specifically designed to provide legal services and based on the latest models from OpenAI.
-
What are the 5 key challenges and the 5 key opportunities raised by artificial intelligence for lawyers in your jurisdiction?
Challenges:
- Job displacement and transformation: AI technologies may lead to the displacement of certain tasks traditionally performed by lawyers, necessitating adaptation and upskilling.
- Interpretation of AI-generated Results: lawyers may face challenges in understanding and interpreting the outputs generated by AI algorithms, requiring a deep understanding of AI technology.
- Cybersecurity and Data Protection: the increased reliance on AI technologies raises concerns regarding the protection of sensitive legal information, requiring robust cybersecurity measures and compliance with data protection regulations.
- Legal liability: lawyers must address responsibility and accountability issues, including allocating legal liability among human operators, AI developers, and AI systems.
- Ethics and Discrimination – bias and trust: AI raises complex ethical and legal questions, such as the responsibility and accountability of AI systems and the potential biases and discrimination they may exhibit.
Opportunities:
- Enhanced Efficiency and Productivity: AI can automate repetitive and time-consuming tasks, enabling lawyers to focus on more complex and strategic aspects of their work.
- Data Analysis and Predictive Analytics: AI enables lawyers to analyze vast amounts of legal data, extract valuable insights, and make informed predictions, enhancing their decision-making capabilities and providing a competitive advantage.
- Document Automation and Contract Analysis: AI-powered tools can help automating the drafting of legal documents, saving time and reducing certain human errors.
- Legal Research and Due Diligence: AI-powered research tools can quickly sift through vast legal databases, helping lawyers in conducting comprehensive legal research and due diligence tasks more efficiently.
- Innovation and new Practice Areas: AI creates opportunities for lawyers to explore new areas of practice, including legal technology consulting, AI policy and ethics, compliance, and intellectual property rights concerning AI.
-
Where do you see the most significant legal developments in artificial intelligence in your jurisdiction in the next 12 months?
France closely follows European legislative developments on AI and actively participates in the creation of harmonized European legislation. The aim is twofold: to protect European citizens better and to bolster confidence in AI, stimulate investment, and foster innovation.
The highly anticipated draft AI Act aligns with the trend of increased operator accountability to prevent potential harm. It adopts a risk-based approach, prohibiting certain AI applications and subjecting “high-risk” systems to compliance reviews and specific requirements before market release. Following the adoption of the draft by the European Parliament in June 2023, the trilogue period has begun and negotiators aim to reach a final agreement by the end of 2023.
To ensure the effectiveness of preventive measures and establish a consistent normative framework, the European Commission also proposed two directives in September 2022 for liability rules related to AI-caused damages.
The first proposal extends the scope of the 1985 Directive on defective products to include the digital economy and AI systems. All economic operators, including producers and entities in the supply chain, would be held responsible. Objective liability remains, without the need to prove fault, but new criteria tailored to AI systems are introduced.
The second proposal establishes a specific liability regime for AI systems, complementing the draft AI Act. Its purpose is to harmonize rules for claims beyond the Product Liability Directive and address specific issues like privacy violations or damages caused by security breaches. The draft directive also aims at simplifying the burden of proof and establish the causal link. Victims, both individuals and companies, are protected under this liability mechanism, regardless of their professional capacity.
France: Artificial Intelligence
This country-specific Q&A provides an overview of Artificial Intelligence laws and regulations applicable in France.
-
What are your countries legal definitions of “artificial intelligence”?
-
Has your country developed a national strategy for artificial intelligence?
-
Has your country implemented rules or guidelines (including voluntary standards and ethical principles) on artificial intelligence? If so, please provide a brief overview of said rules or guidelines. If no rules on artificial intelligence are in force in your jurisdiction, please (i) provide a short overview of the existing laws that potentially could be applied to artificial intelligence and the use of artificial intelligence, (ii) briefly outline the main difficulties in interpreting such existing laws to suit the peculiarities of artificial intelligence, and (iii) summarize any draft laws, or legislative initiatives, on artificial intelligence.
-
Which rules apply to defective artificial intelligence systems, i.e. artificial intelligence systems that do not provide the safety that the public at large is entitled to expect?
-
Please describe any civil and criminal liability rules that may apply in case of damages caused by artificial intelligence systems.
-
Who is responsible for any harm caused by an AI system? And how is the liability allocated between the developer, the user and the victim?
-
What burden of proof will have to be satisfied for the victim of the damage to obtain compensation?
-
Is the use of artificial intelligence insured and/or insurable in your jurisdiction?
-
Can artificial intelligence be named an inventor in a patent application filed in your jurisdiction?
-
Do images generated by and/or with artificial intelligence benefit from copyright protection in your jurisdiction? If so, who is the authorship attributed to?
-
What are the main issues to consider when using artificial intelligence systems in the workplace?
-
What privacy issues arise from the use of artificial intelligence?
-
What are the rules applicable to the use of personal data to train artificial intelligence systems?
-
Have the privacy authorities of your jurisdiction issued guidelines on artificial intelligence?
-
Have the privacy authorities of your jurisdiction discussed cases involving artificial intelligence?
-
Have your national courts already managed cases involving artificial intelligence?
-
Does your country have a regulator or authority responsible for supervising the use and development of artificial intelligence?
-
How would you define the use of artificial intelligence by businesses in your jurisdiction? Is it widespread or limited?
-
Is artificial intelligence being used in the legal sector, by lawyers and/or in-house counsels? If so, how?
-
What are the 5 key challenges and the 5 key opportunities raised by artificial intelligence for lawyers in your jurisdiction?
-
Where do you see the most significant legal developments in artificial intelligence in your jurisdiction in the next 12 months?