US threatens Anthropic with deadline in dispute on AI safeguards

US Secretary of Defense Pete Hegseth vowed to remove Anthropic from his agency's supply chain if the company declined to allow its artificial intelligence (AI) technology to be used across military applications.

The threat was issued on Tuesday at a Pentagon meeting that Hegseth had demanded with Anthropic boss Dario Amodei, a source familiar with discussions told the BBC.

"We continued good-faith conversations about our usage policy to ensure Anthropic can continue to support the government's national security mission in line with what our models can reliably and responsibly do," Anthropic said in a statement.

A senior Pentagon official said Anthropic had until Friday evening to comply.

A source told the BBC the tone of the discussion between Hegseth and Amodei was cordial, but Amodei laid out what Anthropic considers to be its red lines.

These include involvement in autonomous kinetic operations in which AI tools make final military targeting decisions without human intervention.

The use of Anthropic tools for mass domestic surveillance constitutes another red line, the source said.

But the Pentagon official told the BBC the current conflict between the agency and Anthropic is unrelated to the use of autonomous weapons or mass surveillance.

The official said that if Anthropic did not comply, Hegseth would ensure the Defense Production Act was invoked on the company.

That measure could compel Anthropic executives to allow unrestricted use by the Pentagon on national security grounds.

The official added that the Pentagon would simultaneously label Anthropic as a supply chain risk.

An Anthropic spokesperson said Amodei "expressed appreciation for the Department's work and thanked the Secretary for his service" during the meeting with Hegseth.

Anthropic is the maker of the AI chatbot Claude and was one of four AI companies to be awarded contracts with the Pentagon last summer.

Google, ChatGPT-maker OpenAI and Elon Musk's xAI which makes the AI chatbot Grok were also awarded contracts of up to $200m (£148m) each.

Defence department official Emil Michael has previously said the agency wants OpenAI, Google, xAI, and Anthropic to allow the Pentagon to "be able to use any model for all lawful use cases."

Anthropic has consistently aimed to position itself as a more safety-orientated approach to AI research as compared to rivals.

It regularly shares safety reports on its own products with the public.

One such report from last year acknowledged its AI technology had been "weaponised" by hackers who used it to conduct sophisticated cyber-attacks.

The company's image was challenged after reports that the US military used its AI model Claude during the operation that led to the capture former Venezuelan President Nicolás Maduro in January.