The European Union (EU) has achieved a major agreement, set to become one of the broadest regulations of artificial intelligence (AI) in the Western world. The agreement strives to find a balance between promoting innovation and safeguarding the rights of both individuals and companies, according to Thierry Breton, the EU’s internal market chief.
As reported by Bloomberg,
he emphasized the EU’s intention to eschew mass surveillance.
After more than 37 hours of discussion, the European Commission, the European Parliament, and representatives from 27 member countries concluded on a series of controls for generative AI tools. These tools, such as OpenAI Inc.’s ChatGPT and Google’s Bard, can generate content on demand.
The agreement permits limited live facial scanning but includes safeguards and exceptions. The deal explicitly outlaws biometric scanning that categorizes people based on sensitive characteristics, like political or religious beliefs, sexual orientation, or race. Violating companies can face fines of up to €35 million or 7% of global turnover, depending on the nature of the violation and the company’s size.
Although the draft legislation still requires formal approval from EU member states and the parliament, it represents a substantial progression towards establishing an influential AI policy. In the absence of substantial actions from the US Congress, the EU is leading the way in setting robust boundaries for AI technology outside of Asia.
Key points of contention in the discussions revolved around the regulation of AI startups in France, like Mistral AI, as well as Germany’s Aleph Alpha, and the potential societal risks they pose. France and Germany were resistant to regulations they viewed as unnecessary restraints on local businesses.
Agreements were largely reached on rules around generative AI, which include basic transparency requirements for any developer of a large language model. Those posing systemic risks will need to adhere to a voluntary code of conduct to work with the commission to mitigate risks.
The use of live biometric identification tools in public spaces emerged as a significantly contentious issue. The parliament initially voted for an all-out ban last spring, yet, EU countries sought exemptions for national security and law enforcement. The final agreement restricts the use of such technology in public spaces with additional precautions.
Approval is still needed from the parliament and the Council, which represents the EU’s 27 member states. Both France and Germany had previously expressed concerns that the act might overly regulate general purpose AI models and stifle domestic competitors like Mistral and Aleph Alpha.