Key Highlights

  • The U.S. government ordered Anthropic to disable Claude Fable 5 and Mythos 5 worldwide, citing national-security concerns over a jailbreak that surfaced software vulnerabilities. Anthropic is complying but publicly disputes that a narrow, non-universal jailbreak should justify recalling consumer models — a story dissected at length by both Theo (t3.gg) and the All-In Podcast.
  • A separate developer backlash over Fable 5’s safeguards — mandatory 30-day prompt retention (even for zero-retention enterprise customers) and silent model “downgrading” when frontier-AI research is detected — fueled accusations of surveillance and anti-competitive behavior. Anthropic walked back part of it, agreeing to disclose downgrades.
  • NVIDIA topped the first agentic-AI infrastructure benchmark (Artificial Analysis AgentPerf), with GB300 NVL72 running up to 20x more agents per megawatt than Hopper.
  • Google sued “Outsider Enterprise,” a China-based AI-powered phishing operation tied to 9,000 fake sites and 2.5M scam texts in two weeks, while pushing seven bipartisan anti-scam bills.
  • Business moves: Jeff Bezos unveiled Prometheus (an “artificial general engineer,” $12B raise at $41B), and Mistral is rumored to be raising €3B at a €20B valuation.

Analysis & Opinion

Anthropic’s safety warnings may have just backfired — the government has pulled the plug on its most powerful AI — TechCrunch

The U.S. government, citing national-security authorities, ordered Anthropic to immediately disable Claude Fable 5 and Mythos 5 for all users worldwide — a directive received Friday at 5:21pm ET. Mythos, Anthropic’s most capable model, had been previewed in April and restricted to ~50 vetted organizations (Amazon, Apple, Microsoft) via “Project Glasswing” because of its ability to find software vulnerabilities; Fable 5 was the public, guardrailed version released just three days earlier. The government’s concern reportedly stems from a demonstrated jailbreak that asked the model to read a codebase and “fix any flaws,” which can be reverse-engineered into vulnerability discovery. Anthropic counters that the flaws found were minor and that other public models — including OpenAI’s GPT-5.5 — can find the same issues without any bypass. The episode is a striking real-world test of frontier-model safety governance: Anthropic argues that recalling a model over a narrow, non-universal jailbreak would, if applied industry-wide, “halt all new model deployment.” Other Anthropic models remain available.

How we’re combatting AI scams with security, legislation and more — Google The Keyword

Google filed suit to dismantle “Outsider Enterprise,” a China-based, Telegram-operated cybercrime network selling subscription phishing kits ($88/week) that impersonate banks, telecoms, and retailers. The scale is large: 9,000 fake sites, over 1 million fraudulent URLs, and 2.5 million malicious texts sent to Android users in two weeks — with hundreds of thousands of victims and losses in the millions. Google is coordinating with the FBI and carriers to block the texts, advocating for seven bipartisan anti-scam bills, and leaning on AI-powered Android defenses that block over 10 billion malicious messages monthly. (TechCrunch covered the same lawsuit.)

Meta’s months-old AI unit is a ‘soul-crushing gulag,’ say the engineers stuck inside it — TechCrunch

Meta’s three-month-old Applied AI team — ~6,500 reassigned engineers and PMs, many describing themselves as “draftees” who learned of the move via surprise emails — is mired in morale problems, including a heckler disrupting a livestreamed company presentation. Their core task: generating puzzles and coding problems to train AI models, work some call “soul-crushing.”

Andrew Yang thinks the next big startup opportunity is lowering the cost of living — TechCrunch

Yang argues the next wave of startups should cut what Americans overpay on essentials — housing, education, food, fuel, transport, media, and wireless — modeled on Mark Cuban’s Cost Plus Drugs. His own MVNO, Noble Mobile (launched Sept 2025), returns rebates for unused data and has reached thousands of customers and millions in revenue.

Developers are emotionally attached to their tools — Stack Overflow Blog

A look at why developers form strong emotional bonds with their editors, languages, and toolchains — and what that loyalty means as AI-assisted tooling reshapes daily workflows.

New Products & Tools

Jeff Bezos’ $41B ‘artificial general engineer’ — Rundown

Bezos unveiled Prometheus, an AI startup co-founded with physicist/chemist Vik Bajaj, raising $12B at a $41B valuation to build an “artificial general engineer” for complex physical machinery — aiming to compress decade-long design cycles (e.g., a 10% jet-engine thrust gain). Bezos framed AI as creating “more than 10x” new opportunities rather than displacing jobs.

Mistral is rumored to be raising €3B at a €20B valuation — TechCrunch

Per Bloomberg, the French lab is in early talks for ~€3B ($3.5B), nearly doubling its €11.7B valuation from last September. Mistral continues its open-weights-plus-specialized-closed-model strategy as Europe’s leading frontier lab.

Deploy Long-Context Reasoning and Agentic Workflows with MiniMax M3 — NVIDIA Developer Blog

MiniMax M3 is a 428B-parameter MoE model with native multimodality and up to 1M-token context. Its MiniMax Sparse Attention (MSA) pre-filters relevant context instead of full quadratic attention, delivering 4x+ faster KV-cache access at million-token scales.

What’s New in WeatherMesh-6 — WindBorne Systems

WindBorne’s WM-6 is billed as the most skillful medium-range forecasting model to date (NWP or AI), at 25km resolution. Its 4.5-day forecast reportedly matches IFS’s 1-day accuracy, and it adds 24 new surface/atmospheric parameters plus a redesigned ensemble data-assimilation architecture.

Research

NVIDIA Blackwell Leads on First Agentic AI Infrastructure Benchmark — NVIDIA

On Artificial Analysis’s new AgentPerf (AA-AgentPerf) benchmark, GB300 NVL72 ran up to 20x more agents per megawatt than Hopper. The benchmark measures concurrent agents sustained under model-specific SLO tiers (normalized per accelerator and per megawatt) using private, prerecorded coding-agent trajectories spanning 12+ languages and 5K–131K-token sequences — capturing the chained LLM-plus-tool-call patterns that distinguish agentic from conversational workloads. (Also covered on the NVIDIA Developer Blog.)

Interviews & Conversations

BREAKING: Fable and Mythos have been taken down for security concerns — Theo - t3.gg (14 min)

Transcript-based summary. Theo walks through Anthropic’s official statement on the government export-control directive that suspended all access to Fable 5 and Mythos 5 — including for foreign-national Anthropic employees. He sides largely with Anthropic: the demonstrated jailbreak (asking the model to read a codebase and fix flaws) surfaced only minor, widely-discoverable vulnerabilities, and Anthropic argues GPT-5.5 can do the same — raising fears that GPT-5.5 could be next. He highlights Anthropic’s “defense-in-depth” posture — strong safeguards, mandatory 30-day prompt retention to research jailbreaks, and active monitoring — and ties the moment to Dario Amodei’s freshly relevant essay on “the AI exponential,” which argues policy is moving roughly a year behind AI’s pace. Theo also floats that the action may be retaliatory, coming days after Anthropic urged the government not to block state AI laws absent federal standards.

Anthropic’s Fable Backlash, Nationalizing AI, Inflation Heats Up & California’s Broken Elections — All-In Podcast (102 min)

Transcript-based summary (AI segment). The hosts dig into the Fable 5 backlash, calling Anthropic’s safeguards a turning point: Chamath warns that pre-screening prompts and silently deciding what output you “deserve” creates censorship risk for individuals and a non-starter governance risk for enterprises. David Sacks frames it as the “regulatory capture campaign based on fear-mongering” he predicted eight months ago, citing mandatory 30-day retention (overriding zero-retention enterprise agreements) and silent model “nerfing” — with overreach examples like users downgraded for asking about mitochondria or GLP-1 drugs. Friedberg, running a genomics company, says tightening bio-restrictions are pushing real scientific work toward locally-run open-source models — and the best open models today are Chinese, which he argues undermines U.S. competitiveness. The panel debates whether free-market competition (rather than a proposed FDA/FAA-style model approval agency) is the answer, and steelmans Dario Amodei’s view that the same capabilities that cure cancer also enable cyber, physical, and bio weapons — drawing a Manhattan Project analogy and arguing regulation should target weaponized outputs, not the tools themselves. Chamath notes the enormous capital barrier (a gigawatt of compute now ~$100B) to building an open-source alternative at scale.

References

  1. Anthropic’s safety warnings may have just backfired — the government has pulled the plug on its most powerful AI — TechCrunch, 2026-06-13 [blog]
  2. How we’re combatting AI scams with security, legislation and more — Google The Keyword, 2026-06-12 [blog]
  3. Chinese cybercrime operation that used AI to scam hundreds of thousands of victims sued by Google — TechCrunch, 2026-06-12 [blog]
  4. Meta’s months-old AI unit is a ‘soul-crushing gulag,’ say the engineers stuck inside it — TechCrunch, 2026-06-12 [blog]
  5. Andrew Yang thinks the next big startup opportunity is lowering the cost of living — TechCrunch, 2026-06-13 [blog]
  6. Developers are emotionally attached to their tools — Stack Overflow Blog, 2026-06-12 [blog]
  7. Jeff Bezos’ $41B ‘artificial general engineer’ — Rundown, 2026-06-12 [blog]
  8. Mistral is rumored to be raising €3B at a €20B valuation — TechCrunch, 2026-06-12 [blog]
  9. Deploy Long-Context Reasoning and Agentic Workflows with MiniMax M3 — NVIDIA Developer Blog, 2026-06-12 [blog]
  10. What’s New in WeatherMesh-6 — WindBorne Systems, 2026-06-12 [blog]
  11. NVIDIA Blackwell Leads on First Agentic AI Infrastructure Benchmark — NVIDIA, 2026-06-12 [blog]
  12. NVIDIA Achieves Leading Agentic Coding Performance on First Agentic AI Benchmark — NVIDIA Developer Blog, 2026-06-12 [blog]
  13. BREAKING: Fable and Mythos have been taken down for security concerns — Theo - t3.gg, 2026-06-13 [video]
  14. Anthropic’s Fable Backlash, Nationalizing AI, Inflation Heats Up & California’s Broken Elections — All-In Podcast, 2026-06-13 [video]