EU running in circles trying to get AI Act out the door
Bloc risks missing out on first-to-legislate status if timetable slips
The European Commission is entering 11th-hour negotiations to hammer out legislation governing the development and use of AI in time to retain its position as first mover on AI rules.
Talks between EU member states are set to start this afternoon and could go well into the night as representatives struggle to resolve their differences and clear a path to form the basis for further negotiations.
Points of contention remain on whether the law should govern how foundational models are developed in machine learning and whether AI can be used for live biometrics monitoring such as facial recognition.
Forrester principal analyst Enza Iannopollo said: "The potential delay on the adoption of the EU AI ACT is bad news. In fact, the opposite is what one wishes for. The time to press on the accelerator on AI regulation is now. AI is evolving fast and it's already in the hands of millions of users.
"Risks are real and we must address them quickly if we are to benefit from this technology. While discussions are happening in many countries about AI rules and safeguards, the European Union is best placed to produce effective change in the form of a risk-based, principle-led regulation, which will significantly influence the global scenario and inform future policy decisions.
"Undue delays on the adoption of the EU AI Act, however, threaten its ability to be an effective tool for mitigating AI risks and to produce the change that we, as a society, need urgently."
But without a deal, the Commission is likely to postpone legislation owing to a lack of time, sacrificing the EU's position leading the world in regulating AI.
- AI won't take your job, might shrink your wages, European Central Bank reckons
- Getty's image-scraping sueball against Stability AI will go to trial in the UK
- Boffins devise 'universal backdoor' for image models to cause AI hallucinations
- Microsoft fixes Copilot multi-monitor issues in Windows 11 update
Alexandra van Huffelen, Dutch minister for digitalization, told Reuters: "The world is watching us: citizens, stakeholders, NGOs and the private sector want us to agree on a meaningful piece of legislation regarding AI, including general-purpose AI."
She said it was vital that national representatives reached a compromise on general-purpose AI to allow negotiations to proceed.
Others said taking the time to get the legislation right would be equally important. Trying to align a number of fundamental issues might be difficult in the time remaining.
"The AI regulation as a whole is not yet such that one can speak of being ready for a decision," said German digital minister Volker Wissing at the meeting of the European communications and transport ministers.
He said there was still no agreement, even on central definitions, which needed to be resolved before the legislation could proceed.
While the EU strives to progress plans for legislation first mooted two years ago, the tech industry continues to introduce generative AI in new product releases across the board.
Questions remain over the provenance of the training data – some of which may be copyrighted material – and the reliability of the answers.
Meanwhile, Big Tech stalwarts such as IBM, Meta, Intel, Red Hat and Oracle rolled out their AI Alliance yesterday, seemingly trying to act on standards as a unit, with Linux Foundation, NASA and academia all having a seat at the table. However, as we pointed out yesterday, OpenAI and its backer, Microsoft, as well as the maker of the GPU grunt behind a lot of AI processing, Nvidia, are going their own way.
While the EU wants to lead the world in law, other jurisdictions such as the UK and US are set to adopt a lighter touch to regulating AI. ®