Europe to vote on AI laws with potential 7% revenue fines
Risk-based approach puts onus on developers and sees bloc diverge from US, UK plans
The European Parliament is to debate legislation that could cement a divergent approach to regulating AI between the US, the UK and the EU.
Proposals set to go before the legislative chamber in the world's richest economic trading bloc this week include a ban on real-time remote biometric identification systems in public spaces, and biometric categorisation systems that place people in groups by gender, race, and ethnicity, for example, predictive policing and emotional recognition systems.
The Register understands the focus on generative AI, such as OpenAI's ChatGPT, has climbed the agenda since the laws were first proposed.
After it comes into force, the Act could see fines of €40 million or 7 percent of the total worldwide turnover imposed on organizations failing to comply.
But as well as prohibited applications, the legislation looks to put an onus on AI developers and users to ensure they meet the requirements of a risk-based approach markedly different from the UK's proposed outcome-based approach. The US has so far failed to come up with federal legislative proposals specific to AI, although the White House has published a blueprint for an AI Bill of Rights intended to "guide for a society."
The EU's proposed legislation sets out a definition of "high risk" in terms of health and safety or fundamental rights. Developers of systems deemed high risk will have to comply with "a set of horizontal mandatory requirements for trustworthy AI and follow conformity assessment procedures before those systems can be placed on the Union market."
The Commission is set to establish a system for registering standalone high-risk AI applications in a public EU-wide database. Meanwhile, for all AI, the Act sets out an onus on developers to assess of the quantity and suitability of training data sets, as well as examine them for possible biases.
Academics have raised concerns that publicly available training datasets contain biases at the very least, and in some cases, quite horrific content.
The Register understands the EU hopes the regulations will stimulate the creation of high-quality training data sets with commensurate economic value.
UK goes its own way
While the EU pursues a risk-based approach to AI legislation, the UK has set out on a different path with its outcome-based approach.
"There are no outright bans proposed for the UK, no detailed compliance requirements on data, documentation, testing and certification, no new regulator and no new powers to impose fines or take action. Instead, the UK is planning a ‘bottom up’ approach, devolved to regulators," law firm Osborne Clarke explained in a blog.
- AI, extinction, nuclear war, pandemics ... That's expert open letter bingo
- Tough Euro crackdown on AI use passes key vote
- EU legislates disclosure of copyright data used to train AI
- EU lawmakers fear general purpose AI like ChatGPT has already outsmarted regulators
The UK's consultation on its proposed legislation comes to a close on 21 June.
Post-Brexit Britain is optimistic about advocating a "proportionate approach that promotes growth and innovation."
Following his meeting with US president Joe Biden, UK prime minister Rishi Sunak announced plans for the island nation to host the first global summit on the regulation of AI this autumn. The idea is that "like-minded allies and companies" would work to develop an international framework to ensure the safe and reliable development and use of AI.
The European Parliament is set to debate its legislation on Tuesday and vote on it on Wednesday. ®