June 29, 2026

4 Minute Read or Listen ->

🔱 Sign up for a free demo today!

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

THIS WEEK'S EDITION

The White House Has Come For ChatGPT

Over two weeks ago, The White House shut down Anthropic's latest AI models. Then last week, it came for OpenAI. The new ChatGPT model, GPT-5.6, will only be available to customers that the Administration directly approves. In today's short read, we explain the de facto AI licensing regime that now exists, how markets are responding, and an explosive new requirement that every AI user in the U.S. could be subject to any day.

The De Facto Licensing Regime 

As of the time of this writing, here's where things stand:

🔱 Anthropic's Fable 5 is still dark. Launched June 9 as the most powerful model ever released to the public, it was pulled three days later by a Commerce export-control letter and remains offline — even after Commerce let the underlying Mythos 5 back for ~100 approved cyber-defense firms on Friday.

🔱 GPT-5.6 is blocked. OpenAI's newest ChatGPT model will go live for only a handful of institutions that The White House signs off on.

As one executive told Politico, it amounts to "a de facto European-style licensing regime." So the same Silicon Valley that heavily bankrolled this Administration largely to keep AI un-regulated just got the opposite.

But It's Much Worse Than European-Style Regulation

Here's the problem with that analogy to European-style regulation. In the EU, regulation may be onerous, but it is at least highly transparent. The current state of U.S. regulation is anything but. There is no AI law granting the White House this power, so Commerce reached for the only lever it had: export controls — the machinery built to keep missile parts from adversaries. And the process runs completely in the dark: Anthropic got its directive on a Friday at 5:21 p.m. with 90 minutes to comply and no specifics on the threat. Over two weeks later, Fable is still down for reasons that no one has put on the record — no criteria, no appeal, no timeline. More than 75 cybersecurity experts wrote to Commerce Secretary Howard Lutnick demanding an "open, scientific and transparent" process. For now, that process seems to consist of text messages between Lutnick and AI executives.

The Market Fallout

Until last week, both OpenAI and Anthropic were planning two of the largest IPOs in history for later this year. The timing of both IPOs is now in doubt, and the Nasdaq just fell nearly 5% in a week. Remember that the entire AI data center buildout, the largest infrastructure buildout in American history, is premised on the expected compute demand from AGI (artificial general intelligence). But with all new frontier models subject to a White House permission slip, it's unclear when, if ever, all those data centers will be necessary.

Coming Any Day: A Biometric Scan To Use AI?

Current rumors are that Fable may return as soon as this week — but with a big catch. The current Administration Order bars "any foreign national" from using Fable, and for the last few weeks, Anthropic has been quietly building the machinery to comply: a scheduled July 8 policy update allows Anthropic to require a government photo ID, a live selfie, and a biometric face scan, handled by a vendor backed by Peter Thiel — the Silicon Valley investor who bankrolled Trump's 2016 run and mentored and funded Vice President JD Vance. If Anthropic proceeds with this policy, expect its employees (including many non-U.S. citizen engineers) to quit en masse, dramatically destabilizing the most advanced American AI company, and the sector more broadly. In fact, many Anthropic engineers would have no choice, because the law would ban them from working on the very models they build.

Further, there is no reason to believe the identity verification will be limited to Fable or ChatGPT-5.6. Current models on the market from both Anthropic and OpenAI have the same alleged security flaws that the Administration claims underpin the export ban for Fable. So it's entirely possible that sometime in the coming days, you will be required to verify your citizenship before using even the chatbots on your desktop.

What This Means For Public Affairs

Regular readers know where this leads: every move against the U.S. frontier labs hands momentum to open-weight models — the kind you download and run yourself, with no off switch anyone can flip. We've tracked that shift for weeks here, and it just accelerated.

🔱 The downloads tell the story. Google's freely available Gemma 4 has reportedly blown past 200 million downloads since its April launch, with the broader Gemma family now topping 400 million.

🔱 The frontier gap has nearly closed. China's GLM-5.2 now matches or beats OpenAI's GPT-5.5 on several coding benchmarks at roughly one-sixth the cost, and lands within a point of Anthropic's best model on a closely watched test.  

🔱 Enterprises are bringing AI in-house. A Databricks survey found more than 75% of large companies now run two or more model families. The largest organizations are standing up their own GPU clusters to fine-tune open-weight models on proprietary data. Given the extreme regulatory instability, a model that can't be revoked is a critical hedge for any organization deploying AI.

Public affairs organizations are no different. Below are two things that every public affairs team should do immediately:

Two Ways To Operationalize This Email

🔱 Talk to HR. If access to frontier models could soon require proving citizenship, know which of your people — and your vendors' — have a verification issue before an arbitrary rule cuts them off in the middle of one of your critical workflows.

🔱 Stress-test your closed-model dependencies. Any closed model in your stack can now go dark by government order with no warning. Know which workflows break, and have an open-source fallback ready.

🔱 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.