A standoff is unfolding between the Pentagon and one of the country’s leading artificial intelligence companies.
According to Axios, War Secretary Pete Hegseth has given Anthropic CEO Dario Amodei until Friday evening to agree to terms that would remove key safety guardrails from the company’s Claude AI model — or face potential government retaliation.
At issue is how far the U.S. military can push commercial AI systems.
Pentagon officials reportedly want Claude available for “all lawful purposes,” including classified and potentially lethal applications. Anthropic has resisted expanding those permissions, citing ethical concerns over fully autonomous weapons and domestic surveillance.
Reports indicate that Hegseth is seeking:
- AI-assisted targeting without direct human control
- Expanded surveillance capabilities
- Removal of what he has described as “ideological constraints” that limit lawful operations
Officials have warned that failure to comply could result in contract termination, designation as a “supply chain risk,” or possible invocation of the Defense Production Act.
Anthropic’s Claude remains the only major AI system currently cleared for use on classified military networks, heightening the stakes. Other firms — including Google, OpenAI, and xAI — have reportedly agreed to broader terms that would give the Pentagon unfettered access to their technologies.
The clash highlights a widening divide between safety advocates who call for the implementation of AI guardrails and defense officials who argue that national security requires fewer operational restrictions.
READ NEXT: Urban Conflict Prompts New Questions About Border Safety
