Google has updated its contract with the U.S. Defense Department to allow the agency to use the company's commercial artificial intelligence systems for classified tasks, a company spokesperson confirmed.
The amendment grants the Pentagon API access to Google’s commercial models, enabling a direct technical connection between Department of Defense systems and Google’s software. The spokesperson emphasized that the arrangement does not include custom work or bespoke model development commissioned by the Pentagon.
"We believe that providing API access to our commercial models, including on Google infrastructure, with industry-standard practices and terms, represents a responsible approach to supporting national security," the company spokesperson said.
In the same statement, Google reiterated a limitation on certain uses of its technology, noting the companys continued support for established public and private sector norms.
According to the spokesperson, the company "remain committed to the private and public sector consensus that AI should not be used for domestic mass surveillance or autonomous weaponry without appropriate human oversight."
The contract amendment marks a new stage in Googles engagement with the Pentagon on artificial intelligence. The companys decision to allow API-based access follows public pressure from the research community: on Monday, several hundred AI researchers sent a letter to Alphabets chief executive urging him to decline requests to enable classified workloads for U.S. defense missions.
The public statements accompanying the contract change do not disclose implementation timelines, technical specifications of the API access, or financial terms. The companys confirmation focused on the nature of access - commercial models via APIs - and on the exclusion of custom model development from the amended agreement.
Context and immediate implications
By limiting the scope to API access for commercial models and explicitly ruling out tailored model-building, Google frames the arrangement as a procurement of its standard services rather than a bespoke defense technology collaboration. The companys reference to industry-standard practices and terms suggests a reliance on existing contractual and infrastructure safeguards.
At the same time, the public appeal from hundreds of AI researchers underscores active concern within the research community about the use of commercial AI systems for classified defense operations, a concern that remains salient following the contract amendment.
What the article does not specify
- Any financial details or pricing for the API access.
- Technical specifications, security controls, or timelines for deployment.
- Whether the amended contract includes limits on the types of classified workloads beyond the exclusion of custom model development.