Five Major AI Labs Agree to Government Checkpoint Before Shipping New Models

Every major U.S. frontier AI lab has now agreed to submit models for federal evaluation before public release. Google, Microsoft, and xAI signed new agreements Tuesday with the Commerce Department's Center for AI Standards and Innovation (CAISI). OpenAI and Anthropic renegotiated existing deals to align with Trump's AI Action Plan.

CAISI operates within NIST and was established in 2023 as the AI Safety Institute, then renamed last June under the Trump administration. Commerce Secretary Howard Lutnick framed the rebrand as moving away from regulation described as "national security" cover. The center's mandate remains: evaluate frontier models for cybersecurity, biosecurity, and chemical weapons risks. It has completed more than 40 model assessments, including unreleased state-of-the-art systems.

CAISI director Chris Fall described the scope: "These expanded industry collaborations help us scale our work in the public interest at a critical moment." Fall's predecessor, Collin Burns — a former Anthropic and OpenAI researcher — left after four days. The Washington Post reported White House officials saw Anthropic ties as a conflict.

For enterprise AI architects, pre-release government evaluation is now baseline practice. Organizations building AI governance frameworks should treat CAISI assessments as an emerging reference point for vendor risk — not a safety guarantee, but a documented evaluation layer.

The timing carries weight. One day before these agreements, reports surfaced that the Trump administration was weighing a mandatory pre-release review process via executive order, citing Anthropic's Mythos model. Voluntary and mandatory frameworks would operate in parallel under such an order. Enterprises in defense or critical infrastructure supply chains may face stricter requirements than voluntary agreements imply.

Anthropic renegotiated its CAISI deal even as the Pentagon's March designation of the company as a supply chain risk remains contested. Two lawsuits are unresolved. Defense Secretary Pete Hegseth and President Trump have outlined a six-month timeline for phasing out Anthropic tools in government use.

CAISI itself lacks permanent legal standing. Draft bills exist to codify the center, but none has passed. Its authority rests on executive direction, not statute. Enterprises treating CAISI clearance as durable regulatory signal should note the legal foundation is contingent on executive stability.

Whether the voluntary framework hardens into statute—or gets superseded by mandatory executive-order review—will define the compliance baseline for next-generation model deployments.