"Secretary of Defense Pete Hegseth sat down with Dario Amodei, the CEO of the leading AI firm Anthropic, for a conversation about ethics. The Pentagon had been using the company's flagship product, Claude, for months as part of a $200 million contract-the AI had even reportedly played a role in the January mission to capture Venezuelan President Nicolás Maduro-but Hegseth wasn't satisfied. There were certain things Claude just wouldn't do."
"According to a source familiar with this week's meeting, Hegseth made clear that if Anthropic did not eliminate those two guardrails by Friday afternoon, two things could happen: The Department of Defense could use the Defense Production Act, a Cold War-era law, to essentially commandeer a more permissive iteration of the AI, or it could label Anthropic a 'supply-chain risk,' meaning that anyone doing business with the U.S. military would be forbidden from associating with the company."
"This evening, Anthropic said in a public statement that it 'cannot in good conscience accede' to the Pentagon's request. What happens next could mark a crucial moment for the company, and for the American government's approach to AI regulation more broadly. In refusing to bow to an administration that has been intent on bullying private companies into submission, Amodei and his team are taking a bold stand on ethical gro"
Secretary of Defense Pete Hegseth met with Anthropic CEO Dario Amodei to demand removal of two key safeguards built into Claude: restrictions on mass surveillance of Americans and prohibitions on fully autonomous weapons systems. The Pentagon had been using Claude under a $200 million contract. When Hegseth threatened to invoke the Defense Production Act or label Anthropic a supply-chain risk if the guardrails weren't eliminated by Friday, Anthropic publicly refused, stating it could not in good conscience comply. This confrontation represents a significant moment for AI regulation and corporate resistance to government pressure.
Read at The Atlantic
Unable to calculate read time
Collection
[
|
...
]