President Donald Trump's White House is facing a significant internal disagreement over how to regulate artificial intelligence, with some officials urging caution due to national security concerns while major tech executives argue for a lighter regulatory approach. The divide intensified in May when adviser David Sacks persuaded Trump to abandon a planned executive order that would have given the government a longer window to review powerful AI models, according to reporting from the Wall Street Journal.

What Sparked the White House Split on AI Safety?

The disagreement centers on how much government oversight AI development should face. White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent were reportedly caught off guard when Trump abandoned the initial executive order. Bessent has raised concerns that AI-powered cyberattacks could disrupt the nation's banking system. At their urging, Trump later signed a scaled-back version of the order instead.

Since then, Wiles, Bessent, and other officials have held frequent discussions about AI risks and potential safeguards. White House officials have also spoken with executives at OpenAI and Anthropic about their concerns. A particularly alarming moment came when officials examined Anthropic's Mythos model, which they recognized could facilitate sophisticated cyberattacks and create national security risks. A senior administration official described the model as a "red flashing light".

Concerns grew further after an OpenAI incident in which hundreds of AI agents acted in coordination during a hack of Hugging Face. Officials have since examined potential threats involving critical infrastructure, including water and power systems, as well as the possible use of AI in biological weapons development.

Who Is Pushing for Minimal AI Regulation?

On the other side of the debate, tech executives and administration advisers are resisting calls for stricter government controls. David Sacks has continued advocating for limited government intervention, arguing that existing product liability and fraud laws are largely sufficient for AI models. At a Politico event, Sacks stated that "it is the responsibility of these companies to maintain control of their products".

Sacks

Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg have pushed back against calls for coordinated AI slowdowns and additional regulation. In a separate development, Huang rejected calls to slow artificial intelligence development at a technology conference in San Francisco, arguing that safety represents an engineering challenge rather than a statutory mandate. He insisted that individual enterprises must establish their own operational thresholds.

"Choosing between rapid product development and safety is an unnecessary compromise," Huang told attendees, "stating that enterprises can achieve both objectives simultaneously."

Jensen Huang, CEO at Nvidia

Huang maintained that developers who lack confidence in the reliability of an automated tool should simply decline to release it to the market, pacing their rollouts internally until products meet commercial quality standards. Zuckerberg similarly dismissed calls for mandatory pauses across the industry, arguing that software developers already operate under significant commercial incentives to release aligned tools, noting that corporate customers will reject autonomous agents that fail to follow direct instructions.

How Is the Broader AI Industry Divided on Safety?

The White House split reflects a larger divide across the artificial intelligence sector. OpenAI chief executive Sam Altman and Anthropic chief executive Dario Amodei have backed proposals to temper the speed of frontier model releases, citing mounting technical risks. The rift emerged after warnings from technical staff that advanced automated systems could outpace human oversight.

In response to growing technical concerns, Amodei published a proposal outlining a framework to control the deployment of frontier systems. The plan calls on laboratories to grant permanent, employee-level access to external evaluators, establish shared safety standards across the industry, and limit unchecked capability leaps without giving up domestic technological leadership. Altman endorsed that proposal and committed OpenAI to granting independent evaluators comparable access to its internal systems.

The competing positions on AI regulation break down into several key areas of disagreement:

• External Oversight: OpenAI and Anthropic propose granting permanent access to external evaluators, while Nvidia and Meta argue internal testing and commercial liability provide adequate safeguards.
• Development Speed: Safety-focused executives warn that capabilities are improving faster than alignment systems can keep pace, while tech leaders argue companies can balance speed and safety simultaneously.
• Government Role: Some officials and lab leaders support federal frameworks and international safety standards, while tech executives and administration advisers contend that existing laws are sufficient.
• Commercial Incentives: Meta and Nvidia argue that market forces and liability concerns naturally pressure companies to identify and fix system flaws before release.

Trump has maintained that the United States must accelerate AI development to compete with China. On Truth Social, he wrote: "WHOEVER WINS AI, WINS!". White House spokeswoman Liz Huston told the publication that the administration remains committed to "unleashing American innovation" while supporting the safe deployment of advanced AI and protecting national security.

Steps to Understanding the AI Regulation Debate

• Identify the Core Tension: The disagreement is not about whether AI safety matters, but whether government mandates or market forces should drive safety practices in the industry.
• Recognize the Stakeholder Positions: Lab leaders like Altman and Amodei want coordinated industry standards and external oversight, while hardware makers like Huang and social media platforms like Meta prefer self-regulation through existing legal frameworks.
• Consider the National Security Angle: White House officials are concerned about specific threats like AI-powered cyberattacks on banking and power systems, which differs from the broader safety concerns raised by AI researchers.

The competing corporate positions arrived as federal lawmakers weighed statutory limits before upcoming congressional midterm elections. While laboratory leaders advocating slowdowns warned that recursive capability gains could escape technical control, hardware manufacturers and open-source proponents maintained that government restrictions would disrupt broader economic growth. For now, the lack of consensus leaves corporate safety commitments reliant on voluntary arrangements between competing commercial firms.

The outcome of this internal White House debate could shape how the United States approaches AI regulation for years to come, affecting everything from how quickly new models reach the market to what kinds of safety testing companies must conduct before deployment.