In partnership with
Hi AI Futurists,
Three weeks ago, the White House ordered Anthropic's best model shut down worldwide over a jailbreak most people outside the industry had never heard of. This week, the ban lifted, right after a very public fight between Anthropic and one of Trump's own AI advisers. Let's take a look.
Our agenda.
Top AI news
Anthropic's fight with the White House just ended
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
HR and IT need to work as one. Here's how
Every missed onboarding step, delayed offboard, or broken provisioning handoff has a root cause: HR and IT aren't aligned. This guide gives both teams a shared framework for the full employee lifecycle.
Top AI News
💻 Meta is opening a business to sell its extra AI computing power (TechCrunch). Meta Compute rents out raw GPUs and access to its Muse Spark model. $182.9 billion committed to infrastructure, and now it's renting out the leftovers.
🎮 General Intuition raised $320 million to teach AI agents from video games (Axios). Khosla led the $2.3 billion round, with Jeff Bezos and Eric Schmidt investing too. The model played Fortnite for 100 hours, then drove a real robot after 8 minutes of data.
🕵️ China-linked hackers are now targeting AI employees, not just networks (CNBC). Insider risk and social engineering are becoming the front line as US-China AI competition intensifies. The firewall problem just became a people problem.
🌍 The UN assembled an AI council of CEOs and heads of state (Axios). Salesforce's Marc Benioff and Rwanda's president co-chair the group, meeting first on July 8 in Geneva. Global AI governance just got its own committee.
💾 AI demand is driving up the price of your next laptop (SiliconANGLE). Micron's revenue quadrupled and SK Hynix filed for a $30 billion IPO as memory chips run short. Prices are projected to climb another 50% by fall, with no relief until 2028.
🛡️ Claude and ChatGPT can now flag a scam link for you (Gizmodo). Norton's "Genie" scam detector plugged into both chatbots this week at no extra cost. Ask either bot about a sketchy link and it checks its own fraud database first.
🤯 A startup raised $100 million to stop AI from contradicting itself (PYMNTS). Scaled Cognition built "APT" after watching models give a different wrong answer to the same question twice. Consistency just became its own product category.
🔬 Anthropic launched a version of Claude built for scientists (MIT Technology Review). Claude Science is free to every paid subscriber and already flagged drug candidates for a rare genetic disease. One Harvard physicist compared its research skills to a sharp second-year grad student.
📉 The world's central bank overseer just called AI spending a bubble risk (Cointelegraph). The Bank for International Settlements flagged over $1 trillion in debt-financed AI buildout as a systemic risk. Even the chip shortage is now showing up in inflation data.
🐉 China built a 1.6 trillion parameter AI model without a single Nvidia chip (SiliconANGLE). Meituan's LongCat-2.0 trained entirely on a 50,000-chip domestic cluster and now competes with Gemini, GPT, and Claude. Export controls were supposed to slow this down.
Anthropic's Fight With the White House Just Ended
On June 12, three days after Anthropic launched Claude Fable 5 and its more powerful sibling Mythos 5, the US Commerce Department ordered both offline worldwide. Commerce Secretary Howard Lutnick's letter cited export-control authority, barring any foreign national, including Anthropic's own non-citizen employees, from touching either model. The trigger was a jailbreak Amazon's own researchers found in Fable 5, a prompt trick that got the model to flag software vulnerabilities and, in one case, produce exploit code. Anthropic pulled both models within days of their debut. For a lab that built its brand on being the safety-first shop, getting shut down by the government over a safety flaw was an awkward way to spend launch week.
The dispute went public fast. White House AI adviser David Sacks, a PayPal alum turned Trump's most visible AI voice, posted that Anthropic had been warned about the jailbreak and refused to fix it or pull the model, writing that the company "prioritized the continued offering of the consumer model over safety." Anthropic pushed back hard, arguing the flaw was narrow rather than universal, similar to capability already sitting in other public models, and that treating it as grounds for a global shutdown would, applied evenly, halt every frontier lab's release schedule. A separate claim that a China-linked group had used the jailbreak never got fully confirmed. On June 26, the government eased the ban for roughly 100 vetted companies and agencies. On July 1, it lifted the order entirely.
Strip away the drama and one precedent survives: the government has now shown it will order a live AI product offline worldwide over a disputed technical claim, with almost no public process. That is not just an Anthropic problem. It is a supply-chain risk for anyone who has built a product on top of a single frontier model. The standoff ran nineteen days end to end and left plenty of unanswered questions about how "serious" a flaw has to be before it triggers a shutdown, and who gets to decide that. The models are back online. The rulebook for next time is still being written, one Friday letter at a time.
Takeaways at a Glance:
On June 12, the Commerce Department ordered Claude Fable 5 and Mythos 5 offline worldwide, three days after launch.
The trigger was a jailbreak in Fable 5 that Amazon's own researchers flagged to the White House.
White House adviser David Sacks said Anthropic "prioritized the continued offering of the consumer model over safety."
On June 26, roughly 100 vetted companies and agencies got access back; on July 1, the government lifted the ban entirely.
It is the first known case of the US government ordering an AI model itself, not the chips it runs on, taken offline.
What We Think About It:
Anthropic's safety-first pitch is part of what got it regulated. A lab that spent two years telling the world how dangerous its models could be shouldn't be shocked when the government takes that claim at face value and acts on it.
The real precedent is the process, not the flaw. A single Friday letter took a live product offline worldwide with no public hearing, and that can happen to any model, from any lab, next time.
What You Can Do Right Now:
If your product runs on one frontier model, line up a second one you could switch to this week, not eventually.
Read the actual timeline before you repeat the headline. "AI broke into government systems" and "a jailbreak triggered an export order" are very different stories.
Investors, watch how this kind of dispute gets resolved next time, not just this time. That process is now part of every frontier lab's risk profile.
Founders, if you're pitching an AI-safety story to press or investors, know that regulators can hold you to your own marketing.
Get more done with these AI tools
Outpaint Ad Reframe: turns vertical social ads into widescreen cuts for TV and streaming
Keeps your original footage untouched; AI only fills in the extra frame space
Turns a reformat around in days instead of weeks of reshoot production
Delivers CTV-ready versions for Netflix, Roku, Hulu, and YouTube alongside your social cut
For marketers repurposing UGC ads across every screen size.
GetCompress: a desktop app that shrinks video, image, and PDF files without the upload
Compresses and converts video, image, GIF, and PDF files locally, no cloud upload required
Cuts file size 50 to 90 percent with barely noticeable quality loss
Watches a folder or your clipboard and compresses automatically in the background
For anyone who emails or uploads large files and hates the wait.
Dotient: a local search app that finds files by what they look like, not what they're named
Searches your files with plain-language, visual descriptions instead of exact filenames
Builds an interactive graph of your archive that clusters similar files automatically
Lets you build "Cubbies," custom collections for mood boards or running to-do lists
For designers, photographers, and anyone whose desktop is a mess of "Untitled_47.psd."
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!