Google amends Pentagon AI deal for classified use

Google has amended a contract with the US Department of Defense to let the Pentagon use its AI models for classified work through API access.

The agreement came as hundreds of Google AI researchers sent CEO Sundar Pichai a letter, urging the company not to make its systems available for classified US defense missions.

They warned that AI can concentrate power and make mistakes.

Google faced similar employee backlash in 2018 over Project Maven, a Pentagon program that used AI to analyze drone footage, and later chose not to renew that contract.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

The contract expands Google Public Sector’s US$200 million Department of Defense AI work

The deal gives the Department of Defense’s Chief Digital and Artificial Intelligence Office (CDAO), the Pentagon group that leads AI adoption, up to US$200 million in work through Google Public Sector. It includes Agentspace and Cloud Tensor Processing Units (TPUs), Google’s AI chips for training and running models 1.
Google Distributed Cloud (GDC), Google’s cloud system for customer-controlled environments, supports the work. Its air-gapped appliance, a system isolated from the public internet, has Department of Defense Impact Level 6 (IL6) authorization for classified workloads 1.
API access to Google’s AI models for classified work would mark a turn from the stance employees pressed for after the 2018 Project Maven dispute. Language that barred military uses was removed from the AI principles in early 2025 2.
A letter signed by 600 Google staff warns that classified settings could keep the company from auditing how its technology is used, which could weaken its responsible AI commitments 3.

The deal marks a new phase for AI in defense

Google’s return to Pentagon work on classified AI suggests large tech firms are growing more willing to act like defense contractors. That shift changes the industry’s ethical debate.
The move also tests employee activism at Google. Staff protests helped end the earlier Project Maven work, yet the company is moving ahead amid fresh internal opposition 2.
Google’s tools could speed military planning and operations, echoing concerns raised during the original Maven debate 4.
Faster development of autonomous systems remains a concern for AI ethicists from that period 4.