Trump Federalizes AI Regulation, Gemini 3 is the Bomb - also, sorry about the empty send earlier!

Sorry about accidentally sending a blank newsletter earlier! Technical difficulties.

AI regulation in the US just went from “chaotic” to “monopoly” in one week.

Trump is about to sign a “One Rule” executive order for AI.
One federal rulebook.
Fifty state rulebooks put on mute.

Read that again.

The same week everyone argued about model benchmarks and cute new agents,
the real game moved to Washington:
Who gets to decide what “safe AI” even means.

Not you.
Not your state.
Not your users.

One rule.
From the top.

Big platforms have been begging for this.
They do not want 50 different AGs breathing down their necks.
They want a single door to knock on, a single set of lobbyists to pay,
and a single battlefield they already know how to win.

If this order lands the way it is being floated, the ripple effects are brutal:

• State-level AI laws get kneecapped before they mature.

• Smaller AI companies lose their best leverage point: local regulators who actually pick up the phone.

• The definition of “responsible AI” gets negotiated in back rooms, then handed to everyone else like a terms-of-service popup.

And here is the plot twist:

At the same time, federal agencies like FDA and HHS are rolling out their own AI strategies and agentic tools.
They are wiring AI into drug review, inspections, workflows, healthcare policy.

So you get:
Centralized rules.
Centralized infrastructure.
Centralized deployment.

Tell me that does not look like an operating system for AI power.

If you build in AI, your reality just changed:

You are not “complying with regulations” anymore.
You are building in somebody else’s jurisdiction, on somebody else’s timeline,
inside a rulebook you did not write.

The winners

• Hyperscalers who can afford permanent residency in DC.

• Giants who love compliance as a moat.

The losers

• Scrappy teams trying to ship novel stuff in health, finance, education.

• States that actually wanted to experiment with stronger guardrails or different risk tolerances.

The spiciest part
Federal preemption does not automatically mean “better safety”.
It can just as easily mean “standardized failure at national scale”.

One bug in the rulebook.
One blind spot in the framework.
One definition of “acceptable risk” that prioritizes GDP over people.

Now that rolls out everywhere at once.

If you are in AI, your job in 2026 is not just “ship features”.
It is:

• Understand the new federal rulebook better than your lawyer.

• Design product strategies that survive when your state’s voice is turned down.

• Decide whether you are playing the game as regulated infrastructure
  or building tools that help everyone else navigate this mess.

Because AI is not just an industry anymore.
It is becoming a federally managed resource.

Treat it like a toy, and you will get treated like a consumer.
Treat it like infrastructure, and you might still have a seat at the table.

If this freaks you out a little, good.
It should.

Because the real AI “alignment problem” right now
is not just aligning models with human values.
It is aligning power with accountability.

And that part is nowhere near solved.