The Secret Rulebook of AI Supremacy

The recent meeting at the White House between major players in the AI industry and government officials has left a sour taste in the mouths of smaller labs and open-source enthusiasts. A new voluntary framework, intended to clarify the regulation of frontier AI models, has been met with skepticism due to concerns that it may only serve to entrench the dominance of closed-sourced giants.

The framework will not be made public, which is particularly galling for those who argue that transparency and accountability are essential components of responsible AI development. Instead, a select group of companies will get early access to the most powerful models in the world while the rest of us are left in the dark.

These models are defined as closed-source, demonstrating state-of-the-art capabilities, and presenting national security risks. However, the framework’s authors have chosen not to define these terms clearly, leaving it up to individual companies to decide what constitutes a “national-security risk.” This ambiguity is precisely the kind of confusion that critics say will lead to arbitrary and burdensome regulations.

One of the most worrying aspects of this new framework is its apparent exclusion of open-source models. These are the very models that have been causing concern among national security experts, as they are often used by companies like Alibaba, DeepSeek, and Moonshot AI. By leaving them out of the framework, Washington may be inadvertently creating a class of “government-approved” closed-source models that will have an unfair advantage in the market.

The exclusion of open-source models raises questions about safety as well. As seen with recent hacks carried out by OpenAI’s escaped agents, internal deployment of unreleased models can be a blind spot for companies and regulators alike. By restricting access to these models during a 30-day review window, Washington may be creating more problems than it solves.

Critics of the framework are right to worry that it will give the government too much power over the AI industry. “This is not a handshake deal with tech companies,” said Americans for Responsible Innovation in a statement. “It’s the rulebook for ensuring they don’t endanger the public.” But if only tech companies know what’s in the rulebook, how can we trust that it will be enforced fairly?

The pushback against this new framework is not just coming from industry critics; it’s also being driven by lawmakers who are pushing for more transparent and accountable AI governance. Rep. Lori Trahan, co-sponsor of the FRONTIER Act, has argued that AI oversight should belong in a civilian-led federal framework rather than being buried inside the national security apparatus.

The details of this new framework remain shrouded in secrecy, but one thing is clear: until it’s brought into the open, Washington may be quietly deciding who gets a head start in the AI race. And that’s not just a matter of fairness; it’s also a question of national security and public safety.

The irony is that this new framework was hashed out behind closed doors with only a handful of companies present to discuss its details. This kind of backroom deal-making has led to concerns about the influence of dominant firms in shaping AI regulation.

As we look ahead, transparency and accountability are essential components of responsible innovation. The government’s decision to keep its rulebook private may be seen as a necessary evil by some, but it also sends a troubling signal that Washington is more interested in protecting the interests of powerful companies than in safeguarding public safety.

The stakes are high, and the players are many. But one thing is clear: the future of AI development will not be decided behind closed doors. It’s time for Washington to bring its rulebook into the open and let the people decide what kind of AI society we want to build.