It’s not about Anthropy vs. OpenAI anymore

The US government is set to assume a tremendous amount of control over the AI ​​models that will be released.

Two weeks after the US government I pulled Anthropic’s Fable and Mythos modelsthe new OpenAI paradigm seems to be heading towards the same oblivion. Information The news broke on Thursday GPT 5.6 will only be released in a limited preview, with the government approving a “client-to-customer” release until a general release is approved.

If this preview only lasts “a couple of weeks,” as Altman predicted, this may not be a particularly big problem. But Mythos has been in preview for several months, and there’s no indication it will be available for general release anytime soon. Even a few weeks of review can dramatically limit the economic upside of an expensive new system, at a time when AI labs are desperately trying to improve their bottom lines. If the pace of model development slows as a result, it will likely lead to a similar slowdown in the ongoing data center buildout process.

If things go wrong, the entire industry could be at risk.

Crucially, OpenAI and Anthropic are now in exactly the same situation with the same problems facing them and the same disaster awaiting them if they fail. Conversations within the tech industry tend to focus on one side or another’s role in making this happen, either accusing Anthropic of running a regulatory takeover scheme or accusing OpenAI of cozying up to Trump to get rid of a competitor. It’s understandable. Many of the most prominent people in the industry own billions of dollars from one company or another.

But what is happening now is bigger than that. The cost of implementing an arbitrary government approval process for each parametric model is clear, and no solution helps one laboratory without helping others.

The most pressing problem is simply creating a logical release process. It’s fine for the government to test models before releasing them (that’s how a lot of consumer products work) — but as GMU fellow (and soon-to-be OpenAI employee) Dean Paul points out… In an eloquent post this morningIt is not clear what kind of safety safeguards could be put in place to satisfy regulators. The US government does not have the expertise or capacity to perform the type of testing that would be needed here. It’s not even clear what regulators will be trying to protect against, as there has been no effort to clarify what risks the government is actually concerned about.

It is tempting to view the government process as the whole problem in itself, but there are real concerns underlying it. Even if you don’t believe the Mythos hype, there’s clear evidence of how AI tools are revolutionizing cybersecurity. There are similar processes at work Biological hazard And alignment. Restricting model releases cannot be the complete solution in itself – this would only limit what is available to the public – but there are real concerns that need to be addressed.

The best ideas for tackling them, as Paul offered, mean working together. This means trusting independent groups to guide the process, even if they don’t exactly align with your goals. This means lining up behind the least bad regulatory options available, rather than fighting every regulation. Most of all, this means fighting for AI as an industry, rather than seeing safety and regulation as an opportunity to gain an advantage.

For many people working in AI, this will be difficult. Unfortunately, however, AI models have evolved to the point where their capabilities have real political consequences. Dealing with these consequences will require collective action. In the coming weeks, we will find out if this is something the industry can do.