What happened
The US Department of Defense awarded Scale AI a $250 million contract on July 11 to run classified evaluation of frontier models under the White House release framework. Scale will operate red-team exercises and benchmark suites in secure environments for models submitted by Anthropic, Google, OpenAI and others.
The contract runs through August 2027. Scale said it will use its existing SEAL evaluation infrastructure adapted for classified settings.
Why it matters
Someone has to run the classified benchmarks the White House framework requires, and Scale's SEAL work makes it a natural operator. The contract formalizes evaluation as ongoing infrastructure rather than a one-time exercise.
It also gives Scale a central role in the release-review process that developers will feel indirectly through model availability.
MintedBrain take
You will not see the evaluation results, but Scale's role means the August 1 deadline has real operational backing. Expect the review process to become a recurring step in frontier launch timelines.
Discussion
Sign in to comment. Your account must be at least 1 day old.