The regulation might look like a balanced compromise, but a lengthy and painstaking negotiation still lies ahead
A Regulation three years in the making: The European Commission published yesterday the first legislative effort to regulate artificial intelligence. The framework is the result of three years of work and extensive engagement with industry and academia. A European Strategy on AI was first adopted in 2018, followed by the Guidelines for Trustworthy AI in 2019, and by a White Paper on AI in 2020. With the White Paper, the EC set out its vision for “an ecosystem of excellence and trust” which was referenced by Commissioner Vestager during the press conference.
An approach based on different levels of risk: The key element of the EC’s proposal is the identification of different tiers of risk for AI technologies – the riskier the use cases, the stronger the safeguards that apply to them. ‘Minimal risk’ AI (e.g. video games, spam filters) will go unregulated, whereas ‘limited risk’ applications will face transparency requirements. High-risk technologies (generally those which may interfere with fundamental rights or determine one’s access to work and education opportunities) will be subject to risk assessments and to a set of stringent requirements, such as the need to meet certain standards in the quality of the data they use. Some technologies are classified as ‘unacceptable risk’ AI and will be banned (e.g. ‘social scoring’ systems).
The draft is out, but the path to approval is riddled with obstacles: The AI proposal is bound to spark intense conflict between (and within) European Parliament and European Council. As was the case for previous legislation such as the GDPR, or the e-Privacy proposal (which is yet to be approved), we are already seeing MEPs claiming that the proposal lacks important safeguards – for example on facial recognition, on which 40 MEPs had called for an outright ban on. On the other hand, some member states in the Council are likely to do all they can to make the Regulation less prescriptive. Late in 2020, for instance, 14 Governments led by Denmark called for a more flexible approach based on voluntary labelling schemes. We are in for a lengthy battle.
Source: https://ec.europa.eu/commission/presscorner/detail/en/IP_21_1682
