The Weekly Waypoint, Issue #21
June 15, 2026
It was supposed to be the biggest AI launch of the year. Instead, it became the biggest AI controversy of the decade.
On Monday, June 9, Anthropic unveiled Claude Fable 5 and Claude Mythos 5 — models the company called its most capable ever, exceeding every model they'd previously released to the public. Fable 5 was state-of-the-art on nearly every benchmark tested. Stripe reported it "compressed months of engineering into days." In a 50-million-line Ruby codebase, the model completed a migration in a day that would have taken a human team over two months. It beat Pokémon FireRed using only screenshots — no helper harness, no game-state information, just raw vision. Its protein design work matched skilled human scientists, and it conducted novel genomics research that outperformed models published in Science — while being 100 times smaller.
Anthropic had spent months preparing safeguards. Safety classifiers would catch cybersecurity and bioweapons queries and fall back to the previous Opus 4.8 model. External red-teamers spent over 1,000 hours trying to break the safeguards and found no universal jailbreak. The UK AISI made progress toward one but didn't complete it. Anthropic required 30-day data retention — a real cost to customer trust — specifically to monitor for attacks. By every measure, this was the most carefully guarded frontier model release in history.
It lasted three days.
The Government Steps In
On Friday, June 12, at 5:21 PM ET, the US government issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national — inside or outside the United States, including foreign national Anthropic employees. Because compliance required ensuring no foreign national could access the models, Anthropic had to shut them down for everyone, including all domestic customers.
The government cited "national security authorities" but, according to Anthropic, did not provide specific details. Anthropic's statement revealed that the government appeared to be acting on a report of a narrow, non-universal jailbreak — essentially asking the model to read a codebase and fix software flaws. Anthropic reviewed the same report and found that the vulnerabilities the model identified were minor and "widely available from other models, including OpenAI's GPT-5.5."
Anthropic complied with the order but issued a striking public rebuke: "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."
The company also noted this doesn't adhere to principles they've publicly advocated for — that government blocking should be "transparent, fair, clear, and grounded in technical facts."
This is the second major clash between Anthropic and the US government this year. The Department of Defense previously declared Anthropic a supply chain risk — a label historically reserved for foreign adversaries — and the resulting litigation is still ongoing.
Apple Rebuilds Siri From Scratch
While the AI world was focused on Anthropic's drama, Apple used WWDC26 to make its most aggressive AI move yet. The company introduced Siri AI — not an update, but a complete rebuild powered by next-generation Apple Intelligence architecture.
The new Siri can answer questions about what's on your screen, search across your messages and emails, pull information from the web, and carry on extended conversations. There's a dedicated Siri app for revisiting conversations, synced across devices via iCloud. Visual Intelligence comes to iPad, Mac, and Vision Pro. Siri can draft emails that match your writing style, automatically fix weak passwords by navigating websites on your behalf, and even help split bills via Apple Cash.
Critically, Apple revealed that its Foundation Models were "custom-built in collaboration with Google and its Gemini models" — a partnership that brings Google's AI deeper into Apple's ecosystem while maintaining Apple's privacy-first architecture with on-device processing and Private Cloud Compute.
Siri AI won't be available in the EU on iOS initially, and won't come to China at all while regulatory requirements are worked through. The features are in developer beta now, with public release planned for this fall alongside iOS 27.
Sanders' $7 Trillion AI Sovereign Wealth Fund
Senator Bernie Sanders unveiled legislation that would create a $7 trillion sovereign wealth fund financed by a one-time 50% tax on the stock of the largest AI companies. Any firm doing $200M+ in annual AI sales would be subject to the tax. The fund would generate "hundreds of billions annually" in direct payments to Americans — estimated at $1,000+ per person per year — and fund programs like healthcare, education, and housing.
A seven-member bipartisan Independent Commission for Democratic AI would hold voting shares in taxed companies, giving the public "direct influence over corporate decision-making." The legislation also requires AI firms to split their non-AI businesses from their AI businesses — a provision that could affect companies like xAI, which has merged with X and SpaceX.
Sanders met with Sam Altman, and the two remained "far apart" on how much stake the American public should have. Trump's former AI czar David Sacks called the plan "straight up confiscation of property." With a Republican-controlled Congress, the bill faces steep odds — but Sanders views it as a conversation starter about who benefits from AI.
The Rest of the Week
Meta launched AI Mode on Facebook (June 15), turning years of public posts, Group discussions, Reels, and Marketplace listings into a searchable knowledge base. Meta AI generates conversational answers from social content rather than returning links. The company hasn't said whether users can opt their public posts out, or disclosed accuracy metrics. Given that Facebook Groups contain medical advice from unqualified strangers alongside expert guidance, the misinformation risks are substantial.
China's BAAI unveiled Physis-v0.1 (June 14), described as the world's first general world foundation model. Unlike language models that learn from text, world models learn how the physical world behaves — physical laws, spatial relationships, cause and effect. The Beijing Academy of AI sees this as the next frontier for robotics and embodied intelligence, with humanoid robots demonstrating object manipulation and table tennis at the conference.
OpenAI faces a state attorneys general investigation (June 13), with New York's AG serving a subpoena seeking documents on advertising, user engagement, model sycophancy, consumer data, treatment of minors and seniors. This follows a Florida AG lawsuit and the Tumbler Ridge shooting incident where OpenAI failed to alert law enforcement after flagging the suspect's account.
G7 leaders and AI executives — including Anthropic's Amodei, OpenAI's Altman, and Google DeepMind's Hassabis — called for a US-led coalition to set AI standards, with democratic nations cooperating on frontier AI safety, cybersecurity, and export controls. France urged the US to share cutting-edge AI with allies.
The Takeaway
This week exposed the central tension in AI development: models are now powerful enough that governments consider them national security assets — but the regulatory framework for managing them doesn't exist yet. Anthropic launched with more safeguards than any model before it. The government shut it down anyway, citing concerns that even Anthropic says are matched by competitors' already-available models.
If the standard for pulling a model is "any non-universal jailbreak exists," then no frontier model can survive. The industry needs a governance framework that's transparent, proportionate, and grounded in technical reality. This week showed we don't have one yet.
Meanwhile, Apple is quietly building the AI that most people will actually use — not through benchmarks or safety debates, but through a Siri that can finally do what it was supposed to do a decade ago.
This was the free weekly edition. Paid subscribers get the full deep dive — an 8,000+ word analysis of the Fable 5/Mythos 5 saga, what it means for AI governance, and how the frontier model landscape is shifting under government pressure.