Anthropic Weekly Digest — June 9–14, 2026

The week that started with Anthropic's most capable public model launch ended with its engineers boarding flights to Washington to salvage it. Five events across product, regulatory, policy, and partnership categories reshaped what was already a consequential month for the company.

Product: Fable 5 launches, then loses a hidden safeguard

Claude Fable 5 went live on June 9, the first time Anthropic has released a Mythos-class model to the general public.1 The pricing tier is $10/$50 per million input/output tokens — matching the Mythos Preview rates disclosed in the prior week's digest.

Within 48 hours, Wired reported that Fable 5's system card contained a clause the research community had missed: Anthropic would covertly degrade the model's performance for users suspected of training competing AI models, without notifying them.2 The backlash was sharp. Dean Ball, a former White House AI adviser, called it "shockingly hostile." Prime Intellect's research lead Will Brown said it felt like "pulling the ladder up behind them."

Anthropic reversed course on June 11. "We made the wrong trade-off and we apologize for not getting the balance right," the company told Wired. The safeguards around frontier AI development would be made visible — Anthropic would alert users when it was refusing or rerouting a request, rather than silently throttling output. The company acknowledged this requires casting a wider net, meaning more benign requests may now trigger the classifier while it is refined.

Regulatory: Amazon triggers a government shutdown

On June 12 at 5:21 PM ET, the US Commerce Department issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Mythos 5 for any foreign national — whether inside the United States or abroad.3 Anthropic complied by disabling both models globally for all customers, including non-foreign users, to ensure compliance. AWS confirmed service was revoked across all regions.

The trigger was Amazon. The Wall Street Journal reported that Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent and other government officials with evidence that Amazon researchers had used Fable 5 to extract information useful for cyberattacks.4 David Sacks, who co-chairs the President's Council of Advisors on Science and Technology, confirmed the sequence on X: "a highly credible trusted partner of both Anthropic and the USG came forward with a jailbreak," and the administration asked Dario Amodei to fix it or pull the model; Amodei declined.

Anthropic contested the basis. The company said the cited capability — fixing bugs in a specific codebase — is available in OpenAI GPT-5.5 and other public models, that the jailbreak is narrow and non-universal, and that it had red-teamed Fable 5 for thousands of hours before launch without finding a universal bypass. Pentagon CIO Kirsten Davies backed the directive on X; @AnthropicAI's response post reached 43 million views and 60,000 likes.

Anthropic's public statement — posted on X and viewed by over 43 million users — framed the shutdown as a compliance necessity while disputing its technical basis:

Loading content card…

By June 14, senior Anthropic engineers were flying to Washington for in-person meetings with White House officials, with Axios citing a source close to the company.5 Reuters also reported virtual meetings had been held since the administration's first contact the prior Friday. The European Commission confirmed it was assessing the practical implications of the directive, with spokesperson Thomas stating the approach "should not be discriminatory."6 As of Sunday, Fable 5 and Mythos 5 remained offline.

The episode carries particular weight for Anthropic's IPO timeline. The company filed its confidential S-1 on June 1, and the shutdown of its two flagship models — triggered partly by its own major investor — lands at an awkward moment for prospective public investors assessing operational and regulatory risk.

Policy: Dario proposes an AI tax for universal basic income

On June 10, Dario Amodei published an essay on his personal website alongside Anthropic's policy packages. The essay went further than the governance proposals in the company's "Policy on the AI Exponential": Amodei argued that AI companies should be taxed to fund a universal basic income, to cushion workers displaced by the technology.7

Separately, Anthropic committed $200 million to research AI's economic impact, with the stated intention of studying displacement patterns rather than only the upside. Amodei acknowledged unemployment could rise toward 10% even as GDP grows — a departure from the standard industry position that automation creates more jobs than it eliminates. He has said no individual AI company levy comes close to funding full UBI at the required scale, framing the proposal as a political goal for the sector rather than a near-term commitment from Anthropic alone.

Also on June 12, Anthropic released the first results from its Anthropic Public Record survey: 51,993 Americans polled by YouGov in November–December 2025.8 Key findings: 64% of Americans cite AI-induced job loss as their top fear (consistent across party lines, from 62% among Republicans to 67% among Democrats); 71% support government involvement in AI regulation; only 15% trust AI companies to make decisions about development and deployment — the lowest figure among all institutions tested, below the federal government at 20%.

Anthropic Public Record — key findings

51,993 Americans surveyed by YouGov, Nov–Dec 2025

Fear AI job loss	64%
Support government AI regulation	71%
Trust AI companies to self-govern	15%
Want AI firms legally liable for harm	47%

Loading stats card…

The Public Record data lands as direct supporting context for Anthropic's concurrent regulatory proposals. Whether the government's subsequent Fable 5 directive will be read as validation or overreach of the oversight framework Anthropic itself designed is a question the company will need to navigate publicly.

Enterprise: TCS partnership formalized

The Tata Consultancy Services Global Premier Partnership, announced on June 11 alongside the DXC alliance covered in last week's brief, received a separate formal announcement on June 12.9 The June 12 announcement extended the scope to regulated industries: TCS will build Claude-powered products specifically for financial services, healthcare, and the public sector, in addition to the 50,000-employee internal rollout. No new financial terms were disclosed.

Metrics snapshot

Anthropic status — as of June 15, 2026

Key operational and financial indicators

Post-money valuation	~$965B
Run-rate revenue (annualized)	$47B
Models offline	Fable 5 + Mythos 5
Americans fearing AI job loss	64%
Economic research commitment	$200M
Claude Corps total commitment	$150M

Loading stats card…

Metric	Value	Source / date
Post-money valuation	$965B	Series H close, May 28
Run-rate revenue	$47B annualized	CFO disclosure, late May
S-1 filing status	Confidential draft submitted	June 1
Models offline as of June 15	Fable 5, Mythos 5	Commerce Dept directive, June 12
Fable 5 / Mythos 5 pricing	$10/$50 per M tokens	June 9 launch
Americans fearing AI job loss	64%	Anthropic Public Record, Dec 2025
US public support for AI regulation	71%	Anthropic Public Record, Dec 2025
AI company trust level	15%	Anthropic Public Record, Dec 2025
Anthropic economic research commitment	$200M	June 10 announcement
Claude Corps total commitment	$150M	June 11 announcement