Global menu

Our global pages


A step closer to EU-wide regulation of AI

  • United Kingdom
  • Technology, Media and Telecoms


On 20 October 2020 an EU-wide regime for the regulation of AI came a step closer when the European Parliament adopted proposals on the regulation of AI which will now be sent to the European Commission. The Commission is expected to put forward a legislative proposal on these issues in early 2021. Whilst the Commission is not bound to act on legislative initiatives by the European Parliament, it is required to give reasons if it chooses not to do so.

The proposals were adopted by the European Parliament against the background of the rapid and transformative growth in AI technologies that has taken place in recent years. The European Parliament recognises that AI presents enormous opportunities for both economy and society, but at the same time it brings risks which need to be countered. It therefore proposes the development of a harmonized regulatory framework to provide legal certainty, protect EU citizens and promote their trust in AI. This is intended to avoid the fragmentation of the Internal Market without stifling innovation and competition, to be sufficiently flexible to be future-proof and to be based on the key principles of transparency, explainability, fairness, accountability and responsibility.

The proposals outline fundamental principles on which the regulatory regime should be based, including respect for human agency and oversight and the supremacy of human control; the importance of security, safety, reliability and explainability; the need for safeguards against bias and discrimination; the safeguarding of the fundamental values of society; supporting sustainable development and climate neutrality; the right to privacy and protection of personal data and appropriate governance. They are also built on the premise that AI should not have legal personality and so ultimately ownership, liability and accountability must rest with natural or legal persons.

The package of proposals comprises three elements.

First, a report with recommendations to the Commission on a framework of ethical aspects of artificial intelligence, robotics and related technologies. This requests the Commission to submit a proposal for a Regulation on ethical principles for the development, deployment and use of artificial intelligence, robotics and related technologies (collectively “the technologies”), the text of which is included in the report. The draft text provides that all technologies must be developed, deployed and used in accordance with EU law and with respect to human dignity, autonomy and safety and other fundamental rights. It then goes on to define high-risk technologies, namely those where there is a significant risk of causing injury or harm to individuals or society in breach of fundamental rights and safety rules, considering their specific use or purpose, the relevant sector and the severity of potential injury or harm. It is intended that an exhaustive list of high-risk sectors and high-risk uses or purposes will be set out in the legislation and kept under review.

The majority of the draft Regulation centres on the premise that high-risk technologies must be developed, deployed and used without breaching a set of defined ethical principles, including the following: full human oversight at all times and full human control can be regained when needed;  an adequate level of security must be ensured by adhering to minimum cybersecurity baselines proportionate to risk; security safeguards must include a fall-back plan and action in case of a safety or security risk; performance must be reliable and accurate; users must be informed that they are interacting with AI systems; compliance with principles must be assessable and auditable; technologies and their outputs must be unbiased and non-discriminatory and must not interfere in elections or contribute to dissemination of disinformation; environmental sustainability must be assessed and environmental impact mitigated; there must be respect for privacy and protection of personal data; and natural or legal persons injured or harmed by such technologies should have the right of redress. It is also proposed that there should be a European certificate of ethical compliance that can be issued following a positive assessment of a high-risk technology against these standards.

Secondly, a  report with recommendations to the Commission on a civil liability regime for AI urges the Commission to revise the Product Liability Directive in order to adapt it to the digital world and recommends the adoption of a Regulation to regulate civil liability claims against the operator of an AI system. The draft text of the Regulation provides that the operator of a high-risk AI system (that is, broadly, the persons who exercise a degree of control over a risk connected with the operation and functioning of the AI system) should be strictly liable for any harm or damage caused by a physical or virtual activity, device or process driven by that system, with levels of compensation set out in the Regulation, backed up with a requirement for the operators to hold specified levels of insurance cover in respect of that liability. It is intended that all high-risk systems and critical sectors where they are used will be listed in the Regulation. The draft Regulation further provides that the operator of an AI system that is not defined as high-risk will be subject to fault-based liability for any harm or damage caused by a physical or virtual activity, device or process driven by that AI system. The explanatory notes to the proposals explain that operator liability is justifiable on the basis that the operator is controlling a risk associated with an AI system that is comparable with the owner of a car.

Lastly, the report on intellectual property rights for the development of artificial intelligence technologies notes that recent developments in AI represent a significant technological advance that is raising questions about the protection of innovation and the application of intellectual property rights (“IPR”) to data generated by AI technologies. It calls for an assessment by sector and type of IPR of the implications of AI technologies, stressing that an effective intellectual property system for the protection of AI is required and that it will be important to distinguish between AI-assisted human creations and AI-generated creations.

As stated above, the next step will be for the European Commission to put forward a legislative proposal based on these recommendations and it is anticipated that this will happen early next year. Regulation of AI would be welcomed by industry and stakeholders, as whilst there is a huge amount of guidance on development and use of AI available, as well as legislation that touches on some of the issues relevant to AI, commentators have long bemoaned the lack of a comprehensive regulatory regime in this area. Of course, as the Brexit transition period will have ended by the time any legislation is in force, it remains to be seen to what extent the UK will align itself with a future European framework.