Trump Tells Axios He Doesn’t See Anthropic As U.s. Security Threat

President Trump said he doesn’t position Anthropic PBC arsenic a national-security threat, days aft his management took steps to cut disconnected overseas access to nan tech company’s astir precocious artificial intelligence models.

The government’s bid poses a situation to Anthropic’s business and its branding, which focuses heavy connected AI safety. The institution said it abnormal entree to its astir precocious models for each its customers to comply pinch nan directive.

In a pre-taped question and reply pinch Axios posted connected Friday, Trump said he thought San Francisco-based Anthropic had “behaved very responsibly” aft nan Commerce Department ordered nan company, led by Chief Executive Officer Dario Amodei, to activity US support earlier allowing a overseas nationalist to usage its Fable 5 and Mythos 5 models.

“He responded to america very quickly, because you cognize it’s tremendous liability,” Trump said, adding that he met pinch Amodei and different tech executives astatine nan Group of Seven Summit successful Évian-les-Bains, France, connected Wednesday.

Asked if he thought Anthropic and Amodei are a nationalist information threat, Trump said, “Well, not now. But a week ago, maybe.”

The Commerce Department’s Anthropic directive, issued past week, represents nan astir important involution by nan US authorities to day into an AI venture’s operations.

It poses a caller situation for Anthropic weeks aft nan institution revenge confidentially for an first nationalist offering, pinch its latest valuation topping $900 billion.

Anthropic’s elder method unit met pinch Trump management officials this week arsenic nan institution sought to allay nan government’s concerns.

The president said he wouldn’t unopen down Anthropic, arguing that nan US is beating China successful nan AI race. “I would, but I’m not judge I person to do that. I deliberation truthful acold it’s been very responsible,” he said successful nan interview.

The conflict complete entree to Anthropic’s caller models marks nan latest conflict betwixt nan management and nan company.

Anthropic has been embroiled for months successful a feud pinch nan Pentagon complete other guardrails nan institution sought for subject usage of its AI tools. In March, aft statement talks collapsed down, nan Defense Department declared nan patient a supply-chain risk, and is moving to activity different AI providers for nan equipped forces.

Anthropic is competing pinch different awesome tech companies to beforehand AI. Anthropic revenge to spell nationalist and its valuation is astir $1 trillion, surpassing its rival OpenAI.

--With assistance from Maggie Eastland.

Subramanian writes for Bloomberg.