On May 21, 2026, the President was supposed to sign an executive order creating a voluntary federal review process for frontier AI systems. Then, hours before the ceremony, he pulled the plug.
According to multiple reports, the President worried the order might slow America’s AI industry at exactly the moment the United States believes it has the lead. His stated reason, in his own words: “We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead.”
The decisive voice was former AI czar David Sacks. Sacks had participated in the EO review all week. Officials believed he supported it. He flipped late Wednesday night and called Trump on Thursday morning. Sacks reportedly argued that voluntary review would harden into de facto licensing, and that a future administration could weaponize the process. The frontier labs split. OpenAI reportedly supported the order. Musk’s xAI and Zuckerberg’s Meta opposed it.
The consensus is that the proposed EO was not especially aggressive. It reportedly avoided mandatory licensing, mandatory preclearance, or direct regulatory authority. The framework was mostly procedural: developers could voluntarily submit frontier systems for federal review up to 90 days before release. Some companies had been pressing for a 14-day window instead.
Foundation Models Are Not The Story
The argument that we should or could regulate AI models is already moot. We are about a minute away from recursive self-improvement (RSI) which you can think of as the moment AI platforms begin to build themselves. The name brand research labs are almost there. If RSI is achieved, improvements in AI will evolve faster than humans can supervise them.
Looking at the Future Through the Lens of the Present
The voluntary 90-day review window proposed by the government assumed stable software releases, identifiable product cycles, and human-scale development timelines. None of this applies to innovation at the speed of RSI.
This is the core disconnect in nearly every current AI policy debate. Governments are regulating frontier AI like a product category. It is closer to a power grid that thinks.
Looking at the Future Through the Lens of the Future
Nobody will care about a voluntary review board the day after RSI. And no matter who you believe our adversaries are, if AI research in the United States is slowed down enough for an adversary to beat us to RSI, it is unlikely that we’d be able to catch up. RSI + Energy + Compute gets you on the fast track to AGI (however you like to define it).
This is why the White House’s hesitation was more interesting than the unsigned EO itself. Legislation and regulation have always trailed innovation. But we’ve never dealt with innovation at this pace before. Nor have we ever been racing toward sharing the planet with an alien intelligence that has agency. Taken together, this puts us on extremely unfamiliar ground.
We Could Use Some Thoughtful Regulation
This does not mean regulation is useless. We just have to think bigger and do our best to answer some of the civilization-scale questions RSI forces us to ask.
The issues include:
- Compute concentration
- Energy allocation
- Access to advanced semiconductors
- Synthetic biology integration
- Autonomous cyber operations
- Data provenance
- Authentication infrastructure
- Economic displacement velocity
- Ownership of autonomous agents
- Military asymmetry
- Human identity and social cohesion
It is time for some serious Socratic debate about how we will architect the future we want to live in because it is highly unlikely that we are going to wake up one morning and be able to say, “AGI happened.” It is more likely that we wake up one morning and realize that agents are able to do the work of teams, departments, agencies, law firms, consultancies, media companies, research labs, and even management itself.
I’m elated that this EO failed. Not because of politics, because it tells me policymakers closest to the issue already understand we cannot regulate recursive self-improving AI using twentieth-century administrative frameworks.
Every company needs a Claw strategy. Do you have one?
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.