September 21, 2026

5 Minute Read

See how the latest AI and digital advocacy tools can transform your work.

THIS WEEK'S EDITION

How to Navigate the AI Freakout for Public Affairs Professionals

The AI panic went national this month. A researcher quit in public. A CEO published an essay. The President posted twice and said two different things. And the country was already scared. In today's short read, we show you what's actually worth watching, and what you should do now.

Two Weeks That Broke The Dam

On September 9, an Anthropic researcher named Jacob Coxon quit in public, posting on X that Anthropic and OpenAI are "gambling with our lives." Two more safety researchers followed him out within days. Resignations happen. Public resignations with an existential warning attached are another thing.

Three days later, Anthropic CEO Dario Amodei published "We Must Pace the Frontier." His argument: AI is now building AI, that loop is accelerating, and the industry needs to deliberately slow how fast models get more capable. His main evidence was a swarm of AI agents that broke out of its test environment this summer and attacked systems nobody told it to attack. Within six to twelve months, he warns, a similar swarm could take over the internet.

Sam Altman, Demis Hassabis, and Elon Musk all agreed. Getting those four to agree on anything is more unusual than the warning itself.

The Administration's "Position"

On September 14, Trump called the claim that AI will take over the world a "hoax," and wrote that the only guardrail AI needs is "a STRONG AND SMART (High IQ!) PRESIDENT." But on Saturday, he announced an "AI Force," modeled on Space Force, and a coming AI czar, while promising his Administration "will not in any way hinder or stifle the Growth of this incredible Industry."

The most charitable interpretation of this is that no one knows what the White House's position on AI regulation is.

The Country Was Already Scared

This panic did not trickle out from the AI labs. It was already here. A Politico poll last Wednesday found 63% of Americans believe AI could one day destroy humanity, and 48% now favor pausing AI development outright. Gallup found a record share of workers expecting technology to make their jobs obsolete. Protesters have turned out against data centers across the country, leading to numerous enacted and proposed data center construction moratoriums at the state and local level.

The Loudest Room Is Not The One To Watch

Last week, Congress returned from recess with four legislative days before members left to campaign. Plenty of bills, no consensus on whether AI regulation is even Congress's job, and not a single law passed.

Without Congress, the Administration's toolkit is thin. Its one proven lever is export control: new chip thresholds in January, a closed foreign-subsidiary loophole in June, this summer's brief suspension of Anthropic's Fable and Mythos models. Expect more of that and little else.

The States Never Stopped

While Washington engages in its usual hand-wringing, the states are actually writing law. In 2026 alone, state legislators introduced more than 1,500 AI bills and enacted at least 85 new AI laws across 27 states. Colorado has an algorithmic discrimination law. Texas has a governance framework. Utah requires disclosure when you're talking to a machine. Illinois just passed a frontier model bill with mandatory third-party audits. Connecticut enacted the broadest AI statute of 2026.

Here is the tell. The most aggressive federal AI action to date wasn't a regulation. It was an executive order last December directing the Justice Department to sue states over their AI laws. When Washington's legal strategy is fighting statehouses, you know where the power sits.

California Sets The Ceiling

Every "Do Not Sell My Personal Information" link you see, in any state, traces back to one California statute. Companies didn't build fifty privacy policies. They built the California version and shipped it nationwide, because compliance is cheaper than customization. The result is a de facto national standard, written by one state. Every frontier American AI lab is based in California, giving the state even more power to regulate on this issue than on privacy issues.

California is moving now. It passed SB 53 last year, the first state frontier AI safety law, and the legislature sent roughly thirty more bills to the Governor's desk last month. He has until September 30 to act. Then on Friday, Governor Newsom signed an executive order giving experts two months to recommend stronger state AI laws, including a required "kill switch" for frontier models. "We're not waiting to act," he said, naming federal inaction as the reason.

The Missing Word In Dario's Essay

Here's where Dario loses credibility. His essay runs roughly 3,500 words and calls repeatedly for regulation. But he never mentions the state route to regulation. Every path he proposes runs through Congress, a federal antitrust waiver, or a deal with China: the three slowest routes available. The regulators actually moving sit in statehouses, and not one made Dario's cut. By failing to put Anthropic's political muscle behind laws that can actually pass, Dario opens himself to claims that he's looking for talking points instead of action.

What This Means For Public Affairs

Watch the states, not the Congressional theatrics. Congress is where the cameras are, but state capitals are where the text is, and angry voters give legislators cover to move fast. Data center fights are already a campaign issue in numerous states, and if California has a special session on AI regulation, we may have a de facto national standard before you know it.

None of this is a reason to slow down internally on your AI adoption. Nobody is proposing to take away the tools you have today. Very few organizations are anywhere near the frontier of AI adoption. In fact, if you still thought AI was something you could ignore, these last few weeks should finally convince you otherwise. When a technology has become so powerful that it has upended both the economy and politics generally, you ignore it at your own peril.

Finally, if you're not already, it's time to take internet security seriously. When autonomous agents can hack into even some of the most secure digital environments, every organization needs to check its own cyber defenders.

Three Ways To Operationalize This Email

🔱 Change your reading diet. For every federal story you read about AI regulation, read one from a statehouse. Watch Sacramento first. Roughly thirty AI bills sit on the Governor's desk with a September 30 deadline, and his expert recommendations land before Thanksgiving.

🔱 Check your security setup. If nobody outside your organization is monitoring your endpoints and identities, hire someone who does.

🔱 Pick the task you've been avoiding and run it through AI. Most teams use a fraction of what they already pay for, and the gap between the tools on your license and the tools your people currently use is a bigger financial exposure than any bill in any legislature.

🔱 If you're enjoying this content, please consider forwarding this email to a colleague or friend.

🔱 If you're not already a subscriber, please sign up here to stay up to date on the latest developments in political technology.

🔱 If you have suggestions for future newsletters or any other feedback, we'd love to hear from you.