In partnership with
Hi AI Futurists,
You may have seen the headline that an AI model broke into the NSA's classified systems in a matter of hours. The real story is stranger: the US government ordered Anthropic to switch off its most powerful model over a single jailbreak, and a week later the president walked the whole thing back. Let's take a look.
Our agenda.
Top AI news
Why the government pulled Anthropic's most powerful AI
3 AI tools to boost your workflow
AI Investment Report
Best,
Lex Sokolin
P.S. Hit reply with suggestions or feedback!
Manage your settings: Share | Unsubscribe | Upgrade
Ecosystem: AI Venture Fund | Fintech Research | Lex Linkedin / Twitter
Six people doing the work. Your headcount is one.
Your finance close runs in #finance. Stripe and QuickBooks reconciled, runway updated, posted Sunday night without you asking.
Engineering review lands in #eng. Viktor pulled the open PRs, left comments on auth-refactor, flagged a dependency blocking api-pagination.
Campaign brief lands in #growth: Meta CPA up 18%, recommendation to pause broad match, a draft landing page already deployed for the variant test.
You hired him on day zero. He lives in Slack and Microsoft Teams alongside your contractors and investors, connects to 3,000+ tools, pushes back when you ship something dumb.
"Viktor is now an integral team member, and after weeks of use we still feel we haven't uncovered the full potential." Patrick, Director, Yarra Web.
Top AI News
🛡️ The US forced Anthropic to pull its most powerful AI (PBS NewsHour). An export-control order took Mythos and Fable 5 offline over a single jailbreak. It only got stranger from there.
📊 Half of US adults now use AI chatbots (Pew Research). Up from a third in 2024, with ChatGPT alone at 44%. Only 29% of users trust what the bots tell them.
🩺 OpenAI's o3 cracked rare-disease cases that stumped doctors for years (NBC News). Working with Boston Children's, it reached 18 new diagnoses across 376 unsolved cases. Every one was confirmed in a lab.
💰 Baseten is raising $1.5 billion for AI's plumbing (TechCrunch). The round values the inference startup at $13 billion, up from $5 billion five months ago. Turns out the pipes pay.
💸 Microsoft may run China's DeepSeek to cut Copilot costs (Gizmodo). Its agents burn so many tokens that flat-rate plans lose money. A cheaper self-hosted model suddenly looks appealing.
⚡ AI data centers got a fast lane to the power grid (TechCrunch). US regulators told grid operators to connect them faster, with the data centers paying their own way. The grid blinked first.
🐕 Anthropic had Claude wire up and run a robot dog (Anthropic). Teams using Claude finished the robotics tasks up to 37 times faster than the group without it. The dog fetched a beach ball.
🚗 GM is dropping Android Auto for its own Gemini system (Engadget). The car itself becomes the assistant, running on Google's AI. Your phone can sit this one out.
🛰️ The AI compute crunch is pushing data centers toward space (CNBC). Google, Nvidia, and a startup called Starcloud are all eyeing orbit, where the sun never sets and the grid never complains.
🇳🇴 Norway is banning AI for young schoolkids (Engadget). No generative AI for ages 6 to 13 starting this fall. The prime minister wants them reading and writing the slow way first.
Why the Government Pulled Anthropic's Most Powerful AI
Here is what actually happened. On June 12, the Commerce Department issued an export-control order barring foreign nationals, including Anthropic's own non-citizen employees, from using two of its models: Mythos, the company's most capable system, and Fable 5, the public version it had released only days before. Anthropic took both offline within about ninety minutes. The trigger was not a break-in. It was a jailbreak of Fable 5, a trick that slips a model past its own safety rules, which Amazon chief executive Andy Jassy flagged to the White House. Mythos is the model behind Anthropic's Project Glasswing, the program that made news in the spring for turning up thousands of long-buried software flaws, some sitting untouched in widely used code for more than a decade.
The NSA part is where the story got loose. A senator, relaying what a general had told him, said Mythos broke into almost all of the government's classified systems not in weeks but in hours. The line went viral, and by June 21 the people closest to it were walking it back. It described an authorized stress test on the government's own networks, the kind where you pay someone to try the locks, not a real intrusion. The more useful development came on June 19, when President Trump told Axios he no longer saw Anthropic as a national security threat. His exact words: "Well, not now, but a week ago, maybe." A week after the most powerful AI model in the country was switched off by federal order, the official line was a shrug.
Strip out the drama and a real precedent is left standing. This is the first time the government has used export controls on an AI model itself, not on the chips it runs on, and it did so with almost no public process. The argument this week moved from whether Mythos is dangerous to whether anyone followed the rules. One former security chief said there were valid findings but nothing unique enough to justify a reaction like this, and others described the government's approach as ad hoc, personalized, and possibly lawless. Access to a model can now end with a Friday order, on the strength of one flagged jailbreak, and the reasoning can flip a week later. The capability headlines will keep coming. The governance question is the one that quietly decides what you are allowed to build on.
Takeaways at a Glance:
On June 12, a Commerce Department export-control order forced Anthropic to take its top model, Mythos, and its new public model, Fable 5, offline.
The trigger was a jailbreak of Fable 5 flagged to the White House by Amazon CEO Andy Jassy, not a security breach.
A viral claim that Mythos "broke into" NSA systems in hours described an authorized stress test, and it was walked back by June 21.
On June 19, President Trump said he no longer viewed Anthropic as a national security threat: "Well, not now, but a week ago, maybe."
As of June 21 the models were still offline, the first time the government has used export controls on an AI model itself.
What We Think About It:
Anthropic's safety pitch is part of what got it regulated. The company built its name on warning how capable and dangerous its models are, and the government took it at its word. Talk up your model as a weapon and do not be shocked when someone treats it like one.
The precedent is the part that lasts. Government can now take a frontier model offline with a Friday order and almost no process, and that is a planning risk for anyone building on one, however this standoff ends.
What You Can Do Right Now:
If your product leans on a single frontier model, line up a fallback you can switch to. This month showed that access can disappear by government order, not just by an outage or a price hike.
Read the actual event, not the viral version. The distance between "AI broke into the NSA" and "a jailbreak got a model pulled" is exactly where panic decisions get made.
Investors, track the process fight, not the capability claims. How this gets resolved sets the rules for every model company in your portfolio.
Watch for the return of Fable 5 and Mythos. The terms they come back under will tell you how far the government intends to shape what these models can do.
Get more done with these AI tools
readywhen: an AI chief of staff that catches every promise you make and drafts the follow-up
Automatically spots the commitments you make across Slack, email, meetings, and Notion
Drafts the briefs, emails, and updates for you to approve, before you ask
Connects in minutes, with nothing to build or configure
For founders and execs who keep dropping their own follow-ups.
Pixlie: an AI video studio that gives you real control over text- and image-to-video
Turns a prompt or a starting image into video, not a black-box guess
Runs a cloud render queue with status, ETAs, and push alerts
Keeps a shared asset library synced across web and iOS
For marketers and creators who want a video workflow they can repeat.
ReleaseDock: AI support chat, a help center, and product changelogs in one widget
Answers from your help center, cites the article it used, and hands off to a human when unsure
Publishes changelogs across the widget, a hosted page, and email
Installs in minutes, no engineering required
For founders and support teams trying to cut their ticket pile.
AI Investment Report
This 158-page research report provides the first comprehensive taxonomy of public companies, private ventures, and tokenized protocols building the infrastructure for autonomous AI systems. Compiled by Lex Sokolin, former Chief Economist at ConsenSys, fintech strategist at Autonomous Research, and current Managing Partner at Generative Ventures, this report delivers institutional-grade analysis of 100+ companies across 14 critical infrastructure layers. Learn more here.
What type of coverage would you like to see most?
That’s all for today, folks!
Reach out to our audience by becoming a sponsor here.
If you’re enjoying the newsletter, share with a friend by sending them this link: 👉 https://www.futureblueprint.xyz/subscribe
Looking for past newsletters? You can find them all here.
Working on a cool A.I. project that you would like us to write about? Reply to this email with details, we’d love to hear from you!