Trump Advisers Signal End to Safety Testing for Open-Weight AI Models, Sources Say

Advisers for a potential second Trump administration have informed leading artificial intelligence companies that it would not conduct safety tests on open-weight AI models, according to a Reuters report citing two sources familiar with the discussions. The communication on Tuesday signals a significant potential shift in U.S. technology policy and the regulatory approach to the rapidly advancing AI sector.

A Potential Policy Reversal

The planned policy would represent a sharp departure from the current administration's stance, which has encouraged voluntary safety commitments from major AI developers. This includes pre-deployment testing, or "red-teaming," for potential risks and societal harms, particularly for the most powerful models.

The focus of the proposed change is on open-weight AI models, also known as open-source models. Unlike proprietary or "closed" systems from companies like OpenAI, the underlying code and parameters of open-weight models are publicly available, allowing any developer to use, modify, and build upon them.

Implications for the AI Sector

A move to end government-led or government-mandated safety testing for these models would likely be viewed favorably by proponents of open-source AI development, such as Meta and France's Mistral AI. These companies argue that an open approach fosters innovation, transparency, and competition.

Ad

However, the policy could raise concerns among those who advocate for stronger safety guardrails, who warn that powerful, unchecked open-source models could be more easily exploited for malicious purposes, such as creating disinformation or cyberweapons. This development creates a clear policy distinction that could impact investment and strategy across the AI industry, favoring companies that champion a less restricted, open-source ecosystem.

Contrasting Regulatory Philosophies

The reported stance aligns with a more hands-off, deregulatory approach to new technology, prioritizing speed and innovation over pre-emptive risk mitigation. It stands in contrast to the Biden administration's executive order on AI and efforts by global regulators to establish comprehensive safety frameworks for advanced AI systems.

For investors, this signals a potential fracturing of the regulatory landscape for AI, a key sector for technology stocks. The policy direction of the next U.S. administration will be a critical factor in shaping the competitive dynamics between developers of open and closed AI models.