🤚 The Open-Palm Executive Summary

After weeks of internal rewrites, industry tantrums, and what sources describe as “aggressive lobbying by people who own buildings named after themselves,” President Trump has signed a new executive order on artificial intelligence that asks — not tells, asks — AI companies to pretty please submit their most powerful models for federal review 30 days before public release.

The original draft demanded 90 days. The industry said that was “too onerous.” The White House apparently agreed, because nothing says regulatory backbone like cutting your timeline by two-thirds because a CEO sent an email.

Key provisions include:

• Section 3 directs NIST to develop a classified benchmarking process for evaluating “advanced cyber capabilities” of AI models
• Establishes “Unbiased AI Principles” requiring models to be “truthful” and avoid “ideological dogmas such as DEI”
• Section 4 prioritizes prosecuting cybercriminals who use AI tools
• A voluntary 30-day review window for powerful new models before public deployment

The word “voluntary” is doing more heavy lifting in this document than any single word has done since “unprecedented” in 2020.

👐 The Two-Handed Policy Dissection

Let us appreciate the exquisite choreography here. The administration needed an AI executive order because everyone has an AI executive order now. The EU has the AI Act. China has its own regulatory framework. Biden had Executive Order 14110, which Trump revoked on Day One of his second term because it had the wrong name on it. And now, after months of operating in a regulatory vacuum that the industry absolutely loved, someone apparently noticed that the vacuum was making a sound.

The 30-day voluntary review is the centerpiece, and it is approximately as enforceable as a library’s “please return books on time” policy. AI companies are requested to submit frontier models to federal review before release. Not required. Requested. Executive orders technically only bind federal agencies — but the implied threat is that if you don’t play along, the government’s $100+ billion in annual technology procurement might find its way to someone who does.

Then there’s the “Unbiased AI Principles” section, which requires models to be “truthful” and free of “ideological dogmas.” Critics were quick to note that the language bears a remarkable resemblance to the marketing copy for Grok, xAI’s chatbot, which has been positioned as the “truth-seeking” alternative to models that occasionally acknowledge the existence of systemic inequality. The executive order, it appears, has a favorite child.

🌿 The Gentle Awakening

There is something almost philosophical about watching the most powerful government on Earth politely ask a handful of companies — most of which are valued higher than the GDP of several NATO members — to slow down for a month before releasing technology that may reshape civilization.

The 90-to-30-day negotiation is the real story. In those 60 lost days lives an entire theory of governance: the state proposes, the market disposes, and what emerges is a document that reads less like regulation and more like a suggestion box that happens to have the presidential seal on it.

Meanwhile, open-source developers are raising a different alarm entirely. If even voluntary reviews create friction, they create disproportionate friction for smaller players who can’t afford a government affairs team. The big labs — OpenAI, Anthropic, Google DeepMind — already have entire departments dedicated to talking to regulators over lunch. The graduate student training a model in a university basement does not.

👑 The Gold-Leaf Reckoning

Here’s what actually matters: this order establishes the principle that the federal government has a role in reviewing AI models before deployment. That principle, however toothless today, is a precedent. Precedents are seeds. Seeds become trees. Trees become lumber. Lumber becomes the regulatory framework that someone in 2029 will call “stifling innovation” while standing on a stage at a conference sponsored by the company being regulated.

The NIST benchmarking provision is the sleeper clause. A classified government process for evaluating AI cyber capabilities means the intelligence community will, for the first time, have a standardized way to assess what frontier models can do when pointed at infrastructure. That’s not nothing. That’s the quiet part.

But the loud part — the “truthful AI” mandate, the voluntary timelines, the conspicuous alignment with one particular billionaire’s product positioning — ensures that the discourse will focus on culture war semantics instead of the genuinely consequential national security provisions buried in the appendix.

Which, if you’re a cynical observer, might be exactly the point.

“The executive order asks AI companies to voluntarily slow down, which is like asking a Ferrari to voluntarily observe the speed limit in a country that just fired all the traffic cops.” — The Slap of Wisdom Regulatory Affairs Bureau, drafting a voluntary compliance letter that it will voluntarily not send