Technology AI, Science and New Things
EU policymakers were locked in last-ditch negotiations through Wednesday night and into Thursday morning on the passing of the EU AI act after a last-minute push by French and German representatives to water down parts of the bill.
The EU AI act would be the world’s first set of comprehensive AI regulations, setting out a risk-based approach that imposes restrictions based on potential harm. It was expected to be passed before 2024, but the rise of generative AI has thrown a spanner in the works – leading European nations are now worried about overregulation as they look to nurture their own nascent AI ecosystems.
In November, France, Germany and Italy signed a paper supporting “mandatory self-regulation through codes of conduct” rather than more onerous restrictions. If the impasse isn’t broken by January, it’s unlikely the bill will be passed before EU parliament elections in June.
The EU is a regulatory rather than technology superpower, and the proposed bill is seen as an effort to force external AI players to conform to EU values. But as the bill reaches its final stages, policymakers are having second thoughts about conceding the technology race entirely to the US and China.
Two AI startups have recently risen to prominence in Europe – Mistral AI in France and Aleph Alpha in Germany. Both have been characterised as European variants of OpenAI; both are building their own language models (Aleph Alpha describes their effort as a “Sovereign” AI model for Germany) and both have close ties to their respective governments – Mistral’s co-founder Cédric O is a former government minister.
At least the EU has options. The UK government now is realistically a rule-taker rather than maker, downstream of decisions taken in Washington and Brussels. There’s no “British” OpenAI – London-based Deepmind was acquired by Google in 2014, and no serious UK AI challenger has emerged since.
