Artificial Intelligence
Pentagon stops using Anthropic AI tools
A US defence official told the BBC the Pentagon had ceased using Anthropic products after months of dispute over military access, safety guardrails and the company’s national-security designation.
- A US defence official told the BBC that the Pentagon had ceased using Anthropic products.
- Pete Hegseth said in February that Anthropic was a supply chain risk and that the Pentagon would stop using it by late August.
- Multiple people familiar with the matter said Claude was still being used as recently as last week for research, analysis, intelligence work and operations against Iran.
- The article said Claude was embedded in Maven Smart System, which is operated by Palantir.
- Anthropic declined to comment on the Pentagon’s statement.
A US defence official told the BBC that the department had stopped using Anthropic’s tools. The article said Pete Hegseth had already described the company as a supply chain risk in February and had announced a phase-out by late August.
The report said it was not clear why the move took longer than planned. It also said several people familiar with the matter believed Claude had still been in use last week for research, analysis, intelligence work and operations against Iran.
Those sources said the model had been treated as a key tool for handling large amounts of information. The article said it had been built into Maven Smart System, the Pentagon platform run by Palantir.
Lauren Kahn of Georgetown’s Center for Security and Emerging Technology said the integration meant the system would have been difficult to change quickly. Anthropic did not comment, while the article said the Pentagon had also signed contracts with Google, xAI and OpenAI.
The BBC said its reporting came from defence contractors, former officials and other people familiar with Pentagon AI use, several of whom were unnamed. It also said Anthropic and Donald Trump’s administration had recently held warmer exchanges, including meetings between Trump and Dario Amodei.
Named in this story
People
- Pete Hegseth
- labelled Anthropic a supply chain risk and set a phase-out deadline
- Lauren Kahn
- said Claude was deeply integrated into Maven
- Dario Amodei
- met Donald Trump at the White House last week
- Donald Trump
- met Dario Amodei and praised him
Companies
- Anthropic
- its AI tools were stopped by the Pentagon
- Palantir
- operated Maven Smart System
- was among the tech companies the Pentagon signed contracts with
- xAI
- was among the tech companies the Pentagon signed contracts with
- OpenAI
- was among the tech companies the Pentagon signed contracts with
Organisations
- Georgetown's Center for Security and Emerging Technology
- was Lauren Kahn's affiliation
Governments and agencies
- US Department of Defence
- said it had ceased using Anthropic products
Products and systems
- Claude
- was reported as still being used by the department until last week
- Maven Smart System
- was the Pentagon data platform where Claude was embedded
- Mythos models
- were said to have been used by Pentagon employees during the dispute
Places
- Iran
- was the country named in reported military operations
How the source tells it
The piece was a straight news report with a tense, adversarial register, driven by a dispute over military AI use and repeated anonymous sourcing.
- conflict It centred on a dispute between the Pentagon and Anthropic over access, deadlines and legal action.
- urgency A fixed phase-out date and claims of use as recently as last week created immediacy.
- alarm National-security language and references to autonomous weapons raised the stakes.