US Government Expands AI Safety Testing for Major Tech Companies

The United States government is taking a more active role in checking the safety of new artificial intelligence (AI) tools. According to BBC Technology, the US Department of Commerce will now test new AI models from major companies including Google, Microsoft, and xAI before these products are released to the public. These companies have agreed to share their technology with the Center for AI Standards and Innovation (CAISI) for evaluation. This process involves testing, research, and setting best practices to ensure that new AI systems are safe and secure.

This new agreement is an expansion of earlier efforts that began under the previous administration. Previously, companies like OpenAI and Anthropic had already agreed to submit their models for similar checks. Chris Fall, the director of CAISI, stated that these collaborations are important for the public interest during this critical time. The center has already conducted 40 evaluations of various AI tools, including some advanced models that have not yet been released to the public. While the center did not name which models were stopped from being released, the testing process is clearly designed to identify potential risks before they reach everyday users.

Each company involved has different AI products that are now under the spotlight. Google’s most famous tool is Gemini, a chatbot that is now being used by US defense and military agencies. Microsoft is well-known for its CoPilot tool. Meanwhile, xAI produces Grok, a chatbot that recently faced public criticism after it was found to create inappropriate images of people. These tools are becoming more powerful, which is why the government wants to ensure they are tested thoroughly.

This new focus on testing marks a change for the current White House. Previously, President Donald Trump signed orders to create an "AI Action Plan." The goal of that plan was to remove "red tape"—a common term for rules and paperwork that slow down business—and to ensure the US leads the world in AI development. The administration wanted to avoid heavy regulation to help companies innovate faster. However, the situation is changing as the US military increases its use of AI. There are growing concerns about how powerful these systems are becoming.

One example of these concerns involves the company Anthropic. According to BBC Technology, Anthropic recently claimed that it developed a model called Mythos that is too powerful to be released to the public. This suggests that some AI systems might be dangerous if they are not controlled correctly. Furthermore, Anthropic is currently involved in a lawsuit with the US Department of Defense. The company has refused to remove its "safety guardrails"—the built-in rules that prevent an AI from doing harmful things—even for government use. Because of these issues, senior members of the White House staff recently met with Anthropic’s CEO, Dario Amodei, to discuss the future of the technology.

Despite these new agreements, the path forward remains complex. Representatives from Google, Microsoft, and SpaceX, which owns xAI, did not provide comments on these new testing requirements. The balance between allowing companies to innovate and keeping the public safe is a difficult challenge. As AI becomes more integrated into military and government operations, the pressure to ensure these systems are reliable will likely continue to grow. The government's decision to expand its testing program shows that it is no longer willing to take a completely "hands-off" approach to this rapidly changing technology. For now, the focus remains on working with private companies to understand the risks of these powerful new tools before they become a permanent part of our daily lives.