Anthropic Watch: Week of June 8–14, 2026

Three days after Anthropic shipped what it called the most capable commercial AI model ever built, the U.S. government ordered it turned off. That is the week in one sentence.

The launch and the shutdown: Fable 5's 72-hour lifespan

Anthropic released Claude Fable 5 on June 9, billing it as the first Mythos-class model safe for general use.1 Priced at $10 per million input tokens and $50 per million output tokens — less than half the cost of Claude Mythos Preview — Fable 5 immediately topped several frontier benchmarks in software engineering, knowledge work, vision, and long-context tasks. Stripe reported that it compressed months of engineering work into a single day. Cursor named it the top model on CursorBench. On June 9 it was the most capable model Anthropic had ever made publicly available.

On June 12, at 5:21 p.m. ET, the Commerce Department delivered an export control directive ordering Anthropic to suspend all foreign-national access to Fable 5 and Mythos 5 — including Anthropic's own non-citizen employees working inside the U.S.2 Because selectively blocking foreign nationals while leaving domestic access intact was operationally infeasible, Anthropic disabled both models for all users worldwide. Every other Claude model remained available.

The trigger was a reported jailbreak. According to Fortune's reporting citing multiple sources, Amazon CEO Andy Jassy raised concerns with senior administration officials on Thursday, June 11, after Amazon researchers used a series of prompts to get the Mythos-class model to provide information about cyberattacks that was supposed to be restricted.3 White House AI adviser David Sacks later confirmed the administration asked Dario Amodei to fix the issue or withdraw the model; Amodei declined, leading to the export control. Semafor also reported, citing unnamed sources, a suspicion that a Chinese-linked group had already used the technique — a claim Anthropic said was never raised in any of its conversations with the government.

Anthropic's public response was unusually direct. The company said it had reviewed what it believes is the basis for the government's directive and confirmed the technique is "widely available from other models including OpenAI's GPT-5.5."2 It also said the government provided "only verbal evidence of a potential narrow, non-universal jailbreak" — essentially showing a model how to read a codebase and flag vulnerabilities, something defenders do routinely. The company described its position: if this standard were applied consistently, "it would essentially halt all new model deployments for all frontier model providers."

The Fable 5 launch page, showing a stylized butterfly graphic — the model's symbol — alongside benchmark comparison data. — Claude Fable 5 launch imagery from Anthropic's blog. 1

The legal and political implications go beyond this one model. Techstrong.ai noted the Commerce Department applied export control law — normally applied to physical goods like semiconductor chips — to a commercial API serving hundreds of millions of users, a "category leap" with no settled legal precedent.4 An administration official told Axios that no other current model on the market poses the same risk — because none crosses the capability threshold Mythos has reached — and that any future model at that level would need government clearance before release.

The reaction outside the U.S. was swift. Canadian Prime Minister Mark Carney warned the episode shows the danger of depending on American AI providers.5 Former French Prime Minister Édouard Philippe said AI is now critical infrastructure, and infrastructure controlled by others can be unplugged. UK hospitals, companies, and researchers had been using Fable 5 before the shutdown.

What to watch: As of Sunday June 14, Anthropic's senior technical staff were in Washington for meetings with White House officials. Anthropic has said it believes the action is a misunderstanding and is working to restore access. The question now is whether Fable 5 returns, under what conditions, and what this precedent means for every frontier model release that follows.

Enterprise deals: TCS, DXC sign as Global Premier partners

While the week's attention concentrated on the shutdown, Anthropic quietly closed two of its largest enterprise deals.

TCS (Tata Consultancy Services) announced on June 11 a global partnership that will put Claude in front of 50,000 TCS employees across 56 countries and build Claude-powered industry products for financial services, healthcare, public sector, and other regulated industries.6 TCS's UK life-and-pensions subsidiary Diligenta, which administers policies for more than 22 million policyholders, will use Claude for customer experience. TCS iON, which runs 75 million assessments per year across 1,500 Indian cities, will provide Claude training and certification. TCS becomes a Claude Partner Network member, using its own operations as "customer zero" before rolling Claude out to client engagements.

DXC Technology (NYSE: DXC) announced the same day a multi-year global alliance, joining TCS as a Global Premier partner.7 DXC's approximately 115,000 employees in 70 countries operate mission-critical technology for the world's largest banks, airlines, insurers, and governments. The company already built DXC OASIS — its AI-native orchestration platform for managed services — using Claude to write over 95% of the code, with an estimated 10x developer productivity gain. OASIS is now in production at more than 50 customers. The alliance will create a dedicated team of Claude-certified forward-deployed engineers (FDEs), recruited from DXC's workforce, certified through Anthropic Academy in 90 days, and embedded directly inside customer environments. Initial domains: insurance, application modernization, cybersecurity operations centers, and legacy code modernization.

The timing is notable. Both announcements landed the day before the government shutdown, leaving two of Anthropic's biggest new enterprise customers scrambling alongside everyone else when Fable 5 and Mythos 5 went dark Friday evening.

Claude Partner Network: June 2026

Scale of the TCS and DXC deals announced June 11

TCS employees getting Claude access

Countries in TCS rollout

DXC workforce (70 countries)

DXC OASIS code written by Claude (%)

통계 카드를 불러오는 중…

Anthropic announced two significant philanthropic commitments this week, totaling $350 million.

Claude Corps ($150M, announced June 11) will hire and embed 1,000 paid fellows in nonprofits across the country for a year, with CodePath managing the initiative.8 Each of the roughly 400 host organizations will receive a $10,000 grant and free Claude credits. Applications are open through July 17, with intentionally broad eligibility — no degree required. Anthropic President Daniela Amodei said the program is meant to help traditional nonprofits and community organizations adopt AI without waiting for the adoption curve to reach them.

Economic displacement fund ($200M, also announced that week) is directed toward research and policy work on workers displaced by AI.8 Daniela Amodei said the company cannot understand what disruption looks like without studying, publishing, and discussing it. These commitments follow the co-founders' pledge to donate 80% of their personal wealth.

Separately, CEO Dario Amodei floated a proposal for an AI-company tax to fund universal basic income during a Fox Business interview on June 11.9 The proposal has not been formalized.

The philanthropic activity is generating its own critics. Bella DeVaan of the Institute for Policy Studies told Fortune: "The fox can't guard the henhouse. They can't be responsible for their own regulation or for their own definition of what their altruistic mandate is."

IPO context: OpenAI files after Anthropic

Anthropic filed its IPO S-1 confidentially with the SEC on June 1 (covered in last week's issue). This week, OpenAI followed on June 9 with its own confidential S-1 filing, one week after Anthropic.10 No pricing, tickers, or timelines have been disclosed by either company. Anthropic's last disclosed valuation was approximately $965 billion, placing it ahead of OpenAI's NYSE-estimated valuation according to multiple reports. Both companies are now formally in the pre-IPO disclosure process, with SpaceX completing its own $75 billion IPO on June 12 — the largest in history — as the backdrop.

The government shutdown of Fable 5 has injected new uncertainty into Anthropic's IPO timeline. An administration official told Axios this week that any future model crossing Mythos-level capabilities would require government review before release. That is a material disclosure risk for any S-1.

Anthropic Public Record: 15% of Americans trust AI companies

On June 12, Anthropic published results from its first Anthropic Public Record — a nationally representative survey of 51,993 Americans conducted November–December 2025.11 Key findings:

Anthropic Public Record — U.S. public attitudes on AI (Nov–Dec 2025, n=51,993)

First wave of Anthropic's national survey series

Fear AI will cause job loss

Support government role in AI regulation

Trust AI companies to govern AI development

Named curing disease as top hope for AI

통계 카드를 불러오는 중…

64% of Americans named job loss as their top AI fear — the leading concern in every state, among both Democrats and Republicans
71% want government involvement in AI development and regulation — a bipartisan supermajority
Only 15% trust AI companies to make decisions about how AI is developed and used — the lowest of any institution tested, below the federal government (20%), state and local governments (19%), and international bodies (20%)
48% named curing disease as one of their top three hopes for AI
47% said holding AI companies legally liable for harm is the highest-leverage action government can take

Anthropic positioned the survey as a baseline for tracking public attitudes as model capabilities advance. The irony of the week's timing — a survey showing 15% public trust in AI companies, released on the same day the government shut down Anthropic's newest model citing national security — is hard to miss.