Google, Microsoft and xAI to give US government pre-release AI model access

Washington DC, May 5. Google, Microsoft, and xAI announced on May 5, 2026, that they will give the US Commerce Department's Center for AI Standards and Innovation (CAISI) pre-release access to their AI models.

After OpenAI and Anthropic

This expands on existing pre-release evaluation arrangements with OpenAI and Anthropic. Now America's five largest AI labs — OpenAI, Anthropic, Google, Microsoft, and xAI — will share models with the government before public release.

Purpose: capability and safety

The aim is to assess model capabilities and improve safety before public release. CAISI will particularly focus on AI's cybersecurity, bio-security, and national security implications.

The impact of Anthropic's Mythos

The move comes after Anthropic's Mythos cybersecurity model triggered US administration interest in oversight. Paul Nakasone, former NSA director, testified to Congress that without government oversight, "a single AI model can shift the national security landscape".

Anthropic's revenue soars

Anthropic's run-rate revenue has now surpassed $30 billion, up from about $9 billion at the end of 2025 — more than triple in seven months. This is one of the fastest revenue growth curves in the industry.

Europe's separate path

The European Union has built its own regulatory framework under the AI Act, which takes a risk-based approach. The European Commission has already requested limited preview access for vetted EU cyber teams to OpenAI's GPT-5.5-Cyber.

India's response

India's Ministry of Electronics and Information Technology said it is "watching the US model carefully". India plans to set up its own AI safety institute by end of 2026, in collaboration with IISc Bangalore.