OpenAI and Anthropic Called for Government-Level Bans on Dangerous AI Models

OpenAI and Anthropic have sent an open letter to the White House's Office of Science and Technology Policy. This is an unusual move: two of the largest AI companies are themselves asking the government to regulate and control them more strictly.

The Core of the Letter and Request

The companies are calling for government restrictions on the development and deployment of AI models that carry certain risks. In essence, OpenAI and Anthropic are asking for the American government to have the right and mechanism to ban developers from deploying new models if they pose a threat to security or national interests.

This is not a request for soft regulation or industry self-governance. The companies are asking for actual bans—where the government can say "no, you cannot release this model." This is similar to how oversight works for nuclear technology, aviation, pharmaceuticals, or food additives: before launching into mass production, there needs to be inspection and approval. This approach ensures that powerful systems don't fall into the hands of those who could use them for harm.

Why They're Proposing This

Dario Amodei, CEO of Anthropic, emphasized that insufficient government control and weak customer vetting could lead to serious threats to society. The company sees the problem in the fact that currently anyone can gain access to a powerful model without sufficient verification of how it will be used, for what purposes, and who exactly will be working with the system.

This is not just about marginal cases or theoretical scenarios. OpenAI and Anthropic are looking at the ecosystem of startups and companies that are growing rapidly and competing with each other in a race for funding and market share. Without government restrictions, this competition could lead to someone releasing a dangerous model without scrutiny, simply to reach the market faster and attract investment. Haste is the enemy of security.

The main risks that concern both companies:

• Using models to create disinformation, manipulation, and deepfakes at a societal scale
• Deployment without verification of customer intent, integrity, and actual intentions
• Lack of mechanisms for penalties, bans, and accountability for violations and data breaches
• Technology leakage into the hands of criminal organizations and hostile states
• Use for automated cyberattacks, phishing, and hacking of critical infrastructure systems

The Paradox of Competition and Barriers to Entry

An interesting economic paradox is at play here. Companies that have invested billions in R&D and achieved global leadership in AI are asking the government to introduce high barriers to entry for competitors. If government approval is required for every new model, small startups simply won't be able to compete—they don't have the resources or connections in Washington to navigate the bureaucratic process. Meanwhile, OpenAI and Anthropic already have relationships with authorities and could negotiate softer conditions or expedited reviews.

But this is not merely transparent business strategy. Amodei, Altman, and the leadership of both companies seemingly genuinely believe that highly powerful AI requires government oversight. If they are correct in this assessment, then regulation would be logical—as it is for aviation, pharmaceuticals, or nuclear energy.

What Comes Next

The White House's response remains unknown. However, the letter arrives at a moment when the U.S. administration is paying more active attention to AI policy and potential risks.

If authorities heed this appeal, the AI industry will change dramatically. Small startups will have to wait months or years for government approval. The process of releasing a new model will slow down. Rapid innovation could become slower. But at the same time, the risks of uncontrolled release of genuinely dangerous systems that could be used for disinformation or harm will decrease.

The alternative is to continue with the current unregulated competition and hope that companies' internal responsibility will be sufficient. History shows that in rapidly growing and promising fields (social media, fintech, autonomous vehicles), such hope often disappoints. The market does not always self-regulate optimally for society.