European lawmakers are still in disagreement over several issues concerning new regulations for artificial intelligence (AI), and any agreement will be postponed until December, according to sources familiar with the matter. The draft rules on AI were planned to be […]
European lawmakers are still in disagreement over several issues concerning new regulations for artificial intelligence (AI), and any agreement will be postponed until December, according to sources familiar with the matter.
The draft rules on AI were planned to be agreed upon between the European Parliament and the member states of the European Union. So far, these drafts have been discussed three times in so-called trilogues, which are meetings between the Parliament and the EU member states to align the final versions of the laws.
The fourth trilogue will take place on Tuesday, one day after EU lawmakers will discuss their negotiating stance on basic models and high-risk AI systems, a source said.
Basic models, such as OpenAI’s ChatGPT, are AI systems trained on large datasets, with the ability to learn from new data to perform various tasks.
Spain, currently holding the EU presidency until December, is pushing for an agreement and has proposed compromises to accelerate the process.
These compromises include a gradual approach to regulating basic models, defined as those with more than 45 million users, according to a draft seen by Reuters.
Spain also wants additional obligations for very capable basic models (VCFMs), like ChatGPT, including regular checks to detect potential vulnerabilities.
Opponents argue that even smaller platforms can be equally risky.
Spain claims to have consulted other EU countries on potential compromises ahead of the fourth trilogue with the EU and the Parliament. However, sources suggest that a final agreement is likely to be reached in that meeting.
The fifth trilogue is scheduled for early December.
If no agreement is reached then, negotiations could be postponed until the beginning of next year. Discussions could then be further disrupted by the European Parliament elections in June.
Several lawmakers, including EU Commissioner for Industry Thierry Breton and co-rapporteurs for the AI Act, Dragoș Tudorache and Brando Benifei, have expressed hope that the draft will be approved before the end of the year.
The EU started working on the AI Act in 2021. In May of this year, the European Parliament adopted a draft legislation that includes new rules regarding the use of facial recognition, biometric surveillance, and other AI applications.
According to the proposals, AI tools will be classified based on their perceived level of risk, from low to unacceptable. Governments and companies using these tools will have different obligations depending on the classification.