Google signs AI deal with the Pentagon, ignoring protest from over 600 employees

The detail that deserves more attention than it's getting is the legal finding that safety provisions in the contract lack enforceable teeth. A contract that includes safety language but cannot be litigated on those terms is, functionally, a contract without safety provisions at all.

This story is largely disconnected from recent activity in our archive, as we have no prior coverage to anchor it to. It belongs, however, to a longer-running structural story about the collapse of the informal norms that briefly governed AI firms' relationships with defense and intelligence clients. Google's earlier withdrawal from Project Maven in 2018 under employee pressure was treated at the time as evidence that internal governance could constrain corporate behavior. This contract suggests the opposite conclusion: that as models become more capable and government contracts more lucrative, the leverage employees hold over these decisions diminishes rather than grows.

Watch whether Microsoft or Amazon move to formalize similar classified-use agreements within the next two quarters. If they do, it confirms that Google's move functions as competitive cover for the rest of the industry rather than an isolated decision.