Microsoft, Google, and xAI commit to US government pre-deployment testing — voluntary AISI evaluation becomes a stacked default

Microsoft, Google, and xAI confirmed they will let the US government test their frontier AI models before public launch — joining Anthropic and OpenAI under the voluntary AISI evaluation regime. The five-lab commitment effectively makes pre-deployment government testing the structural default for any US-headquartered frontier lab, even absent statutory mandate.

The five-lab commitment is the under-noticed convergence of the voluntary-disclosure track and the responsible-scaling track. Anthropic and OpenAI signed up first, framing the move as a safety commitment. Microsoft, Google, and xAI signing on transforms it from individual-lab posture into industry-standard procurement signal — the labs that haven't signed (DeepSeek, Mistral, Meta on Llama 4+) become the conspicuous exceptions rather than the rule.

The Q3 2026 watch is whether this voluntary regime hardens into the actual standard or fragments under the federal-preemption posture. Pre-deployment testing is the kind of policy that's easier to announce than to operationalize — the AISI's bandwidth, the test methodology, and the legal disclosure regime around adverse findings all have to ship before the commitment has teeth.

CNN — Microsoft Google xAI government testing → · Paul Hastings — Trump AI EO summary →