• The Trump administration has informed AI companies that it will not conduct safety testing on open-weight AI models, according to sources familiar with the discussions.
  • The decision marks a potential shift from earlier considerations of increased oversight, signaling a hands-off approach to open-source AI.
  • Industry reactions are mixed, with some praising the move for fostering innovation while others express concerns about potential risks.

A Hands-Off Approach

The Trump administration has told AI companies that it will not safety-test open-weight AI models, according to people familiar with the discussions. The decision, which has not been publicly announced, comes as the administration weighs its stance on AI regulation, balancing innovation with national security concerns.

Open-weight models, which make their underlying parameters publicly available, have been a point of contention among policymakers. Proponents argue that they enable transparency and accelerate research, while critics warn they could be exploited for malicious purposes. The administration's choice to refrain from testing these models suggests a preference for minimal intervention, at least for now.

"The administration is taking a pragmatic approach, recognizing that open-weight models are crucial for maintaining U.S. leadership in AI," said a person familiar with the matter, who spoke on condition of anonymity. "They want to avoid stifling innovation with heavy-handed regulation."

The news comes as several major AI developers, including OpenAI, Anthropic, and Google, have been in discussions with government officials about safety protocols. While the administration has previously signaled support for voluntary commitments, it appears unlikely to mandate federal testing for open-weight models.

Industry Reactions

Reactions within the AI community have been swift. Some researchers and startups that rely on open-weight models have welcomed the decision, arguing that federal oversight would create unnecessary barriers. "This is a win for the open-source ecosystem," said a tech entrepreneur who requested anonymity. "Mandatory safety testing would have been a huge burden for smaller players."

Others, however, have expressed concern about the potential risks. "Without federal oversight, we might see models with dangerous capabilities released without adequate safeguards," said a former government official now working in AI safety. "It's a gamble."

The administration's approach contrasts with that of other countries, particularly the European Union, which has been moving toward stricter AI regulations. The divergence could have implications for global AI governance, as companies may choose to operate in regions with more permissive environments.

Looking Ahead

While the administration has decided not to safety-test open-weight models, it remains unclear whether it will introduce other forms of oversight, such as requiring companies to self-certify or adhere to best practices. Industry observers expect continued lobbying from both sides of the debate.

In the meantime, AI companies are likely to proceed with releasing open-weight models, albeit with increased attention to their potential misuse. The absence of federal testing does not absolve developers of responsibility, and many are expected to implement their own safety measures.

As the landscape evolves, the decision will likely shape the trajectory of AI development in the United States. For now, the message from Washington is clear: when it comes to open-weight models, the administration trusts the industry to self-regulate.