Tech companies Google, Microsoft, and xAI have agreed to participate in a new US government AI testing program aimed at evaluating artificial intelligence tools before their public release. This testing will be conducted by the Department of Commerce’s Center for AI Standards and Innovation (CAISI), which will focus on assessing risks such as cybersecurity and biosecurity associated with AI systems. This initiative marks a shift in approach under President Trump, who previously emphasized reducing regulatory barriers to AI innovation, according to the AI National Policy Framework. Additionally, prior agreements between OpenAI and Anthropic with CAISI have been renegotiated to align with the government’s current priorities in AI oversight.

xAI: xAI is an artificial intelligence company founded to advance scientific discovery through frontier AI systems. xAI has agreed alongside Google and Microsoft to submit its AI models for evaluation by the Center for AI Standards and Innovation under the US Department of Commerce. This step aligns with efforts to assess national security implications of advanced AI.
CAISI: CAISI stands for Center for AI Standards and Innovation, a US Department of Commerce initiative evaluating AI risks and standards. It is now partnering with Google, Microsoft, xAI, and others on pre-release testing of AI models for cybersecurity, biosecurity, and related threats. CAISI has renegotiated prior agreements with OpenAI and Anthropic.
Trump: President Donald Trump leads the United States executive branch and shapes national AI strategy. His administration announced the AI testing program with tech giants, marking a shift from prior deregulation focus to targeted risk evaluations via existing bodies like CAISI. This follows his March AI National Policy Framework promoting innovation acceleration.
Google: Google is a technology company known for developing advanced AI models integrated into products like search and cloud services. Google has agreed to allow the US Department of Commerce’s CAISI to test its new AI tools for risks such as cybersecurity and biosecurity before public release. This participation supports collaborative research and best practice development in commercial AI systems.
OpenAI: OpenAI develops large language models and AI tools focused on general intelligence with safety measures. OpenAI renegotiated its 2024 CAISI agreement and provided early access to ChatGPT5.5 for national security testing, including cybersecurity variants like GPT-5.5-Cyber. The company is also crafting responsible deployment strategies for public sector use.
Anthropic: Anthropic is an AI company prioritizing safety and interpretability in its model development. It entered CAISI evaluation agreements in 2024 under the prior administration, which have since been renegotiated. This continues Anthropic’s involvement in government-led AI risk assessments.
Microsoft: Microsoft is a technology giant that builds AI systems embedded in productivity tools and cloud platforms, including the AI model Copilot. Microsoft has joined the US government’s AI testing program through CAISI to proactively address risks like AI-enabled cyber attacks. The company views these evaluations as key to maintaining security leadership.
Chris Fall: Chris Fall serves as the Director of the Center for AI Standards and Innovation (CAISI). He emphasized the importance of independent measurement science for frontier AI and national security in commenting on the new industry collaborations. His leadership is scaling CAISI’s public-interest work amid rapid AI advancements.
Chris Lehane: Chris Lehane is OpenAI’s chief global affairs officer, handling policy and international relations. He shared on LinkedIn that OpenAI supplied ChatGPT5.5 pre-release for CAISI’s national security evaluations. Lehane highlighted ongoing work on cybersecurity models and public service deployment playbooks.
US Department of Commerce: The US Department of Commerce is a federal executive department tasked with fostering economic development, trade, and technological innovation. It oversees the Center for AI Standards and Innovation (CAISI), which is now testing AI models from major tech firms like Google, Microsoft, and xAI. These evaluations focus on risks including cybersecurity, biosecurity, and chemical weapons.
Center for AI Standards and Innovation: The Center for AI Standards and Innovation (CAISI) operates under the US Department of Commerce to establish standards for AI measurement and risk assessment. CAISI is conducting tests, research, and best practices development on commercial AI systems from companies including Google, Microsoft, and xAI. It has previously evaluated numerous models, including unreleased state-of-the-art ones.

`json
{
“Policy Evolution”: “President Trump’s AI National Policy Framework directs existing domain experts rather than new federal bodies to handle AI oversight.”,
“Risk Focus Areas”: “CAISI targets demonstrable risks in AI such as cybersecurity, biosecurity, and chemical weapons potential.”,
“Precedent Agreements”: “OpenAI and Anthropic’s prior Biden-era pacts with CAISI were renegotiated to fit current priorities.”
}
`