1. White House Forces Fable 5 Offline; Amodei Refuses Ultimatum
Anthropic's Fable 5, released June 9, was pulled offline after the White House flagged deployment partner SK Telecom over suspected China ties, prompting Anthropic to revoke its access; Amazon researchers then disclosed jailbreak vulnerabilities, escalating the situation into a global export-control ban. AI czar David Sacks reportedly told Anthropic to either patch every jailbreak or de-deploy the model, and CEO Dario Amodei refused both. Security experts called the "zero jailbreaks" demand technically impossible for any frontier model, reviving fears of "model sovereignty."
2. GLM-5.2 Lands as the New Open-Weight Favorite
Zhipu AI released GLM-5.2 on June 18, with early testers saying it "passes everyone's vibe check" and positions the open model as competitive with leading proprietary systems. The timing is notable: open-weight players like Zhipu and MiniMax (whose M3 model shipped June 1) are pitching immunity to the kind of government recall that just hit Fable 5.
3. Anthropic Opens Seoul Office as Korean Enterprises Deploy Claude at Scale
NAVER rolled out Claude Code across its entire engineering organization, with Samsung SDS, LG CNS, Nexon, and Hanwha Solutions also announcing large-scale Claude deployments. Anthropic opened a Seoul office and signed an MOU with Korea's Ministry of Science and ICT covering AI safety cooperation and Korean-language model evaluations.
4. OpenAI Acquires Astral, Maker of uv and ruff
OpenAI acquired Astral, the company behind the popular Python tooling uv (package manager) and ruff (linter), folding both into its Codex developer platform. The move deepens OpenAI's push into the software-development workflow as coding remains the most lucrative AI use case.
5. TCS to Deploy Claude Across 600,000+ Employees
Tata Consultancy Services announced a global partnership with Anthropic to roll out Claude across its 600,000-plus workforce and to clients in finance, healthcare, and retail. A parallel deal puts Claude into IT-services firm DXC's systems for regulated industries including banking and airlines.
6. Google to Fund $3.2B New York Data Center Supplying TPUs to Anthropic
Google is putting up $3.2 billion to fund a New York data center that will offer TPU capacity to Anthropic, extending the unusual compute alliance between the two rivals. Separately, US energy regulator FERC approved fast-track (90-day) processing for data-center power requests as the AI build-out strains the grid.
7. Japan's Megabanks Gain Access to Claude Mythos
Japan's government and its three megabanks — MUFG, SMBC, and Mizuho, which collectively manage over $8 trillion in assets — are set to gain access to Claude Mythos, Anthropic's restricted Mythos-class model, through Project Glasswing. It is the first confirmed Glasswing deployment to non-US financial institutions.
8. New Frontier and Open Models Pile Up: Nemotron 3 Ultra, Kimi K2.7, DiffusionGemma
The release cadence stayed relentless: NVIDIA shipped Nemotron 3 Ultra (550B, A55B) on June 4, Moonshot AI released the code-specialized Kimi K2.7 Code on June 12, and Google put out the 26B DiffusionGemma (26B-A4B) on June 10. Fable 5 itself had posted 95% on SWE-bench Verified and 80% on SWE-bench Pro before being pulled offline.
9. Anthropic Publishes First "Public Record" Transparency Report
On June 12 Anthropic released its first Public Record, documenting AI-enabled cyber threats it observed and mitigated across its 2025–2026 model deployments. The report is framed as an ongoing commitment to disclose Claude's real-world impact, both positive and negative — landing the same day the Fable 5 shutdown began.
// KEY TAKEAWAYS
This week's dominant theme is governance colliding with capability: a single government order pulled the world's most capable public model, Fable 5, offline overnight — reviving "model sovereignty" fears and handing open-weight players like Zhipu's GLM-5.2 and MiniMax M3 an instant marketing opening. Meanwhile the enterprise land grab accelerated on every front — Korea, Japan's $8T megabanks, TCS (600k staff), and DXC all deepened Claude commitments, OpenAI bought its way further into developer tooling with Astral, and Google bankrolled $3.2B more compute. The throughline: AI is now infrastructure, and whoever controls the switch — governments or providers — controls the market.