The European Commission is “at the forefront” of artificial intelligence safety and oversight, a Commission spokesperson said on Monday, responding to calls from Anthropic chief executive Dario Amodei for a slowdown in the development of the most advanced AI models.
Commission spokesperson for technology Thomas Regnier said the responsibility ultimately rests with AI companies to determine whether models that are not considered safe should continue to be developed.
“We are, of course, all in favour of having innovative and state-of-the-art models in the European Union, but we do not compromise on our safety standards,” Regnier told accredited journalists.
“It is up to companies to demonstrate that their services are safe for use by our citizens,” he added.
Regnier said the EU has an important role to play in AI safety and oversight and has established the necessary framework to protect citizens in relation to the development and deployment of artificial intelligence.
“I think we can say with quite a lot of certainty that we are at the forefront,” he said.
AI Act provides regulatory framework
Regnier pointed to the EU’s Artificial Intelligence Act, known as the AI Act, which he said is now fully applicable.
The Commission has already sent out its first requests for information under the framework, he said.
Asked whether the Commission supports accelerating or slowing the development of so-called frontier AI models, Regnier said the EU’s position is to support innovation provided it fully complies with the European regulatory framework.
“We are in favour of innovation in Europe. We welcome AI services in Europe, provided, of course, that this is done fully in accordance with the AI Act and the safeguards that we have put in place,” he said.
No known talks with Anthropic
Asked specifically whether Anthropic had contacted the European Commission regarding its proposal to “pace” the development of AI models, Regnier said he was not aware of any such communication.
He added that the Commission would not engage in bilateral discussions of that nature in any case.
Regnier also noted that the EU has access to most of the latest AI models, which are tested, among others, by the European Union’s cybersecurity agency.
“Compliance is not an option. It is an obligation in the European Union,” he said.
Regnier stressed that the Commission takes safeguards for European citizens “extremely seriously”, while maintaining that AI innovation remains welcome when it operates within the EU’s regulatory requirements.
Source: CNA
