Anthropic CEO Back in Pentagon Talks Over AI Safeguards

Anthropic CEO Dario Amodei is back at the negotiating table with the U.S. Defence Department after previously resisting a deal that would drop AI safeguards. The aim is to avoid a 'supply chain risk' designation that could complicate access to federal contracts.

The talks revolve around how Anthropic's AI models can be used by the Pentagon while maintaining guardrails. Anthropic has refused to remove two safeguards from its contract: a ban on using its models for mass domestic surveillance and a prohibition on deploying fully autonomous weapons.

According to a Financial Times report, negotiations involve Emil Michael, the under-secretary of defence for research and engineering, and there is no final agreement yet. Defense officials had threatened the supply chain risk label but had not made it official.

Amodei, in a blog post, argued that AI isn't reliable enough to be trusted without human oversight. The move comes as the broader AI field wrestles with safety, reliability, and governance.

While Anthropic's stance persists, the company could maintain access to its Claude AI tools if a deal is struck, avoiding reputational and contractual penalties. The situation highlights the ongoing tug-of-war between rapid AI deployment and safety safeguards.

#Anthropic #PentagonAI #AIGovernance