Palantir and Booz Allen Move to Restrict External AI Models in Government Work

Palantir Technologies and Booz Allen Hamilton, two of the most prominent technology providers serving the U.S. defence and intelligence community, have reportedly put restrictions in place on the use of large language models from Anthropic and OpenAI, according to a report from Quartz. The move highlights a growing tension for contractors operating at the intersection of commercial artificial intelligence and sensitive government environments: frontier AI tools offer significant productivity gains, but their terms of use, data-handling practices, and compliance postures do not always align with the requirements of defence customers.

For companies handling government data, the calculus around third-party AI models extends beyond capability. Contractors must navigate rules around where data resides, who can access it, and how outputs are logged and audited. Restricting certain commercially available models — even widely adopted ones like those from Anthropic and OpenAI — is one way firms attempt to keep workflows within approved boundaries while they evaluate or build alternatives that meet government accreditation standards.

Nvidia, whose hardware underpins much of the AI infrastructure used across both commercial and public-sector deployments, is also cited in the report as having imposed limits, underscoring that the issue cuts across the technology stack rather than being confined to software providers.

Palantir Technologies, which builds data platforms used by intelligence and defence agencies, saw its shares trade at $170.50, down 0.07% from the prior close of $170.62, valuing the company at roughly $439.7 billion. Booz Allen Hamilton, a major government services contractor, traded at $72.20, up 0.12% on the day from a previous close of $72.11, with a market capitalization near $8.68 billion.

Neither the scope nor the duration of the restrictions is fully detailed in the report, and it remains to be seen whether other primes and subcontractors follow suit as agencies sharpen their guidance on generative AI use in regulated environments.

What to watch

• Any formal policy statements from Palantir or Booz Allen on approved AI toolchains for government programmes.
• Upcoming earnings calls, where management may address AI compliance costs or product roadmaps.
• Additional federal guidance on the use of commercial large language models in contract work.
• Potential rollout of accredited, government-specific AI offerings by the companies involved.

Source: original release