‘This development is a further illustration of why Europe needs to strengthen its technological sovereignty’
Anthropic has suspended access to its latest models, Mythos 5 and Fable 5, for all users after the US government demanded that non-US citizens, both in the US and abroad, should not be able to use the LLMs, citing national security concerns.
The nature of the security concerns has not been specified, and rumours that Chinese hackers had managed to jailbreak the models, and that AWS CEO Andy Jassy contacted the administration, are as yet unsubstantiated.
Anthropic has said it is impossible to deny access to individuals based on their nationality and has therefore disabled the models for all customers.
In a blog post the company pushes back against the US government’s apparent claim that a method exists to bypass (jailbreak) its security guardrails.
It says the government has only provided verbal evidence of a vulnerability. It claims it has invested heavily in safety, with Fable 5 undergoing thousands of hours of red-teaming involving the US government, the UK AI Safety Institute, third-party organisations and internal teams, and that no one has found a universal jailbreak. The company adds that it has implemented a defence-in-depth strategy to reduce the risk that narrow jailbreak methods will be successful.
Anthropic acknowledges the government’s right to block unsafe deployments, but only “as part of a statutory process that is transparent, fair, clear, and grounded in technical facts.”
This action does not adhere to those principles, it claims.
Arguments for sovereignty strengthened
The abrupt shutdown of Anthropic’s latest models has added weight to the arguments of those arguing for greater technological independence from America.
Canadian prime minister Mark Carney said the decision demonstrates the danger of relying on a limited number of US companies. “Nobody has done anything wrong in the situation. But we will have done something wrong if we just accept this, don’t take the lesson, don’t build out and diversify.”
European Commission spokesperson Thomas Regnier said: “This development is a further illustration of why Europe needs to strengthen its technological sovereignty. We are looking closely at the practical consequences of this for European users of these services.”
The move is clearly worrying some US startups, which are concerned that the country is increasingly seen as an unreliable partner.
“This is gonna send shockwaves through every lab and neolab… U.S. export control laws operate under a strict liability standard… they are a very sharp blade,” said Matthew Pines, CEO of AI startup Physical Superintelligence, on X.
Impact on Anthropic’s IPO
Earlier this month Anthropic filed confidentially for an initial public offering (IPO) in the US, days after completing a funding round that valued the business at $965 billion. Whether the shutdown of its newest models will affect its prospects will likely depend on how long the suspension continues.
Longer term, the administration’s actions could have a negative impact on investment in US AI firms because of perceptions of geopolitical risk.
Pentagon CIO, Kirsten Davies, argued on X that the Department of Defense is prioritising national security.
“Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation. America First. Always,” Davies said.
Leave a comment