Pentagon ends use of Anthropic’s Claude tools after safety dispute
The US Pentagon has officially ceased using Anthropic's Claude AI tools, an official confirmed to the BBC, following months of contention with the company over AI safety features. Defence Secretary Pete Hegseth designated Anthropic a national security risk in February after the company refused to remove safety guardrails from its systems, citing concerns about mass surveillance and autonomous weapons. This marked a significant withdrawal from Anthropic, which had been the first advanced AI company to deploy tools in classified Pentagon work since 2024.
Despite Hegseth's August deadline for cessation, Claude remained embedded in Pentagon systems as recently as the previous week, integrated into Maven Smart System (the Pentagon's primary intelligence platform). Experts explained that removing deeply integrated AI tools proved far more challenging than initially anticipated. The Pentagon has since shifted to contracts with Google, xAI, and OpenAI for defence operations.
- Pentagon officially stops Anthropic Claude use after company refused to remove AI safety features
- Claude remained in military systems weeks past the August removal deadline
- Pentagon switching to Google, xAI, and OpenAI contracts for defence operations
New here? Start with this
Claude is an artificial intelligence system created by Anthropic, a technology company. The US Pentagon, which oversees the American military, had begun using Claude to assist with defence and intelligence work since 2024, making Anthropic the first advanced AI company to gain access to this sensitive military work.
Anthropic builds safety features into Claude designed to prevent it being misused for harmful purposes such as mass surveillance or developing autonomous weapons. The Pentagon wanted Anthropic to remove these restrictions so Claude could be used more flexibly for military applications. When Anthropic refused, Defence Secretary Pete Hegseth declared the company a national security risk and ordered the Pentagon to stop using Claude.
The dispute highlights a broader tension between technology companies and the US military over how artificial intelligence should be developed and controlled. The practical challenges became clear when Claude remained embedded in Pentagon systems for months after the official order to cease using it, revealing how difficult it is to remove sophisticated AI tools once they are deeply integrated into important operations.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
The Pentagon's position reflects a legitimate national security concern. Advanced AI tools with safety restrictions may prove insufficient for modern military operations where speed and capability are critical. If hostile nations deploy unrestricted AI systems for military advantage, the United States risks technological disadvantage in a conflict. Anthropic's refusal to work within reasonable military requirements effectively cedes leadership in defence applications to less cautious competitors, potentially weakening national security.
The case against
Anthropic's stance reflects a responsible position on AI safety that becomes more important, not less, when these systems have military applications. Removing safety features from powerful AI tools significantly increases risks of autonomous harm, mass surveillance, and unintended consequences that could endanger civilians and soldiers alike. Maintaining ethical standards in AI development—even under military pressure—is essential for long-term security and global stability. Responsible companies can serve national interests without abandoning safety principles that prevent catastrophic misuse.
AI Business Companies Cricket Sport Technology World
Read the full article at the source →
Originally published by BBC World as “Pentagon stops using Anthropic AI tools months after blacklisting company, BBC told”.