The US Department of Defense has officially ceased the use of Anthropic's AI products, a department official confirmed to the BBC on Monday. This move comes months after Defense Secretary Pete Hegseth designated Anthropic a "national security supply-chain risk" earlier this year, citing the company's refusal to remove safety guardrails from its Claude models for military applications.

The Standoff Over Safety Guardrails

The conflict originated when the Pentagon pressed Anthropic to strip safety constraints from Claude to allow unfettered access for military operations. Anthropic refused, citing ethical concerns regarding mass surveillance and autonomous weapons. In response, Hegseth applied a designation typically reserved for foreign entities posing threats to the US, a move Anthropic labeled "unprecedented and unlawful" and is currently challenging in court.

Timeline and Integration Challenges

Hegseth set a six-month deadline to phase out Anthropic on 27 February, one day before the US and Israel attacked Iran. Despite this order, sources indicate Claude remained in active use until last week. The tool was deeply embedded in the Pentagon's Maven Smart System, operated by Palantir, which handles intelligence, satellite imagery, and drone footage for target identification. While the exact reason for the delay past the late August deadline is not clear, Lauren Kahn, a senior research analyst at Georgetown's Center for Security and Emerging Technology, noted that because Claude was deeply integrated into Maven, removing it would have been difficult. "Once they become integrated it can be painful to remove them," Kahn said, adding that these systems are not "plug and play."

Market Shift Toward Competitors

As the legal battle continues, the Pentagon has pivoted to other major AI providers. Contracts have been signed with Google, xAI, and OpenAI, with OpenAI tools seeing wider adoption in military departments recently. The delay in fully removing Anthropic highlighted the critical role Claude played in sorting vast quantities of data for intelligence gathering and analysis, particularly during military operations against Iran.

Warming Relations with the White House

Despite the friction with the Pentagon, relations between Anthropic and the White House appear to be improving. President Trump met with Anthropic CEO Dario Amodei twice at the White House last week. During one meeting, Amodei stated that while he still had concerns about risks posed by AI, "if we work with the president, we can win safely." Trump subsequently praised Amodei, describing him as "great" and "fantastic."

Key Takeaways

  • Anthropic's Claude was removed from the Pentagon's Maven system due to a refusal to disable safety guardrails.
  • Hegseth set the phase-out deadline on 27 February, but the process extended beyond late August due to integration complexities.
  • The Pentagon is increasingly relying on OpenAI, Google, and xAI for military AI applications.
  • President Trump met with Anthropic CEO Dario Amodei twice last week, signaling warming ties despite the Pentagon's ban.

The Bottom Line

This saga proves that enterprise AI procurement is now a geopolitical and ethical minefield, not just a technical upgrade cycle.