The legal and regulatory landscape surrounding artificial intelligence (AI) has been a point of significant discussion at EU level. Last month the European Parliament added to the debate, publishing the following three legislative proposals:
- a general legal and ethical framework for AI;
- a separate civil liability framework in respect of AI products; and
- proposals aimed at protecting IP rights in the AI context.
To avoid fragmentation and ensure maximum harmonisation across the EU, all three proposals suggest legislating by way of a directly effective Regulation (as opposed to a Directive requiring implementation by member states).
While not directly connected to the Commission’s white paper on AI published in February 2020, these proposals show that the Parliament is keen to make its voice heard in the conversation around developing the law governing AI.
The proposals in more detail
1. A new legal framework for AI
As noted in a previous post, the emergence of AI presents risks and ethical dilemmas, as well as opportunities, for society. The European Parliament has therefore proposed a new Regulation setting out ethical principles to be followed in the development, deployment and use of AI, robotics and related technologies. The proposed Regulation builds on several established principles including “human-centric, human-made and human-controlled” AI and “safety, transparency and accountability”, and would be supported by a specific national supervisory authority in each EU member state. The proposed Regulation imposes more onerous obligations on a fairly broad category of ‘high-risk’ AI, including the introduction of a new certificate of ethical compliance which would be issued by a relevant national supervisory authority.
2. Civil liability: a future-oriented framework
The European Parliament has also proposed a separate framework for civil liability which would sit alongside existing product liability rules in cases where AI systems cause harm to people or property. The proposal seeks to balance protecting the public with encouraging innovation while also maximising legal certainty throughout the chain of liability.
The European Parliament identified a legal gap in the existing liability regime (largely underpinned by the Product Liability Directive), whereby AI systems may cause harm to a wide class of people who have no contractual relationship with the deployer or operator of the AI system and therefore have only a slim chance of receiving compensation for the harm suffered.
As with the first proposal, a distinction is made between:
- ‘high-risk’ AI systems, for which the human deployers would be subject to a strict liability regime (similar to car owners, for example) for any harm or damage caused by a physical or virtual activity, device or process driven by their AI system; and
- other (non-‘high-risk’) AI systems, for which there would be a legal presumption of fault towards the deployer, but no strict liability.
3. IP rights
The European Parliament’s final proposal notes that, despite the importance of IP rights in the development of AI and related technologies, the Commission has not yet fully engaged with the key question of how these rights should be protected.
The report highlights the need for strong protection of AI developers’ IP rights in order to strengthen the international competitiveness of EU-based AI companies and recommends:
- an assessment of patent law in light of the growth of AI technology, exploring the possibility that algorithms and methods used within AI systems may form part of overall technical inventions that qualify for patent protection (see The Lens post on the growth in AI patenting activity in the US here); and
- a consideration of whether copyright in AI-generated works should be granted to the natural person who prepares and publishes it, provided that the designer of the underlying technology has not opposed such use.
4. Next steps
The European Commission is expected to respond to the Parliament’s initiatives early in 2021. It is not obliged to put forward concrete legislative proposals in line with the Parliament’s suggestions, but must give its reasons if it decides not to.
In parallel, the Commission’s own legislative proposals produced following its white paper are also expected in early 2021. While this is later than originally planned, and some reports suggest that the Coronavirus pandemic has diminished the appetite in some member states for tech regulation, it seems that AI regulation is still firmly on the radar of the EU’s key legislators.