White House Briefs AI Firms on Review Plans

The Trump administration has started talks on an executive order that would require government review of new AI models prior to their public launch. This comes after a period of reduced regulations. Last week, White House officials informed representatives from Anthropic, Google, and OpenAI about these potential plans, according to sources familiar with the matter cited in a New York Times report.

The proposed order would create a working group made up of technology executives and government officials. This group would look into oversight methods, such as a formal review process for upcoming AI models. Officials drew inspiration from a British approach, where multiple agencies check if AI systems meet certain safety criteria. A White House spokesperson described the reports as speculation to The Information and noted that any policy updates would come straight from President Trump.

Anthropic's Mythos Model Raises Alarms

Anthropic's Claude Mythos model appears to have prompted this policy discussion. The company unveiled the model in April but chose not to release it to the public. Anthropic stated that Mythos excels at spotting software vulnerabilities, which could lead to a major cybersecurity issue if misused. The National Security Agency already employs Mythos to evaluate weaknesses in U.S. government software.

Anthropic, founded in 2021 by former OpenAI employees including Dario and Daniela Amodei, prioritizes AI safety and alignment. Its Claude series of models has gained attention for strong performance and built-in safeguards. The decision to hold back Mythos highlights ongoing debates about balancing AI advancement with security risks. White House insiders worry about the consequences of an AI-assisted cyberattack causing widespread harm.

Some officials advocate for a system that grants the government early access to new models without halting their rollout. This approach aims to address vulnerabilities proactively.

Shift from Deregulation Policy

This direction contrasts with the administration's initial actions. Upon taking office, President Trump reversed Biden-era rules, such as required safety checks for AI models with possible military uses. In July, Trump referred to AI as a "beautiful baby" that should avoid constraints from "foolish rules" or "even stupid rules."

Vice President JD Vance echoed this at a Paris AI summit, cautioning that too much regulation might harm a key industry. Google, a pioneer in AI through projects like DeepMind and its Gemini models, and OpenAI, creators of the GPT series since 2015, were among those briefed. OpenAI has faced scrutiny over rapid development and safety commitments.

Evolving Political and Personnel Dynamics

Public opinion has played a role in the change. A Pew survey showed 50 percent of Republicans and 51 percent of Democrats feel more worried than enthusiastic about AI's role in everyday life. Bipartisan unease has grown.

Personnel changes also factor in. David Sacks, the AI czar who pushed deregulation, departed in March. Sacks, a tech investor from the PayPal founding group, had influenced the light-touch approach. Now, Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent oversee AI matters. They have indicated to outsiders their intent to guide policy more actively, per the Times.

These developments signal a potential pivot amid rising concerns. The working group could formalize how the government evaluates frontier AI systems from leading labs.

Related on Neura Market

• AI Models Directory
• AI Safety Tools
• Automation Marketplace