TL;DR: The Wall Street Journal reported on June 13, 2026 that the cybersecurity research behind the US export control directive that forced Anthropic to disable Fable 5 and Mythos 5 came from Amazon, and that Amazon CEO Andy Jassy personally shared the company's findings with the White House in the days before the letter landed.[1] The Verge confirmed the WSJ account within hours.[2] Anthropic, which has AWS as its largest cloud partner through Amazon Bedrock, disclosed the directive on the evening of Friday June 12, calling the underlying finding a "misunderstanding" and noting that the same capability is "widely available from other models (including OpenAI's GPT-5.5)."[3] Katie Moussouris, the founder and CEO of LutaSecurity and a co-author of Microsoft's first bug bounty program, posted on BlueSky that she had reviewed Amazon's paper: "It's not a jailbreak."[2] The Hacker News thread on the WSJ story crossed 669 points within a day.[4] This is the political economy of the Anthropic story, and the part that yesterday's day-one coverage could not tell you: a competitor with the ear of the White House used export control authority to remove a rival's most advanced products from the market. The "jailbreak" was the available pretext. The "trigger" was a phone call.
What the Wall Street Journal Reported
The Wall Street Journal story, which broke late June 13, draws on unnamed US officials and people familiar with the conversations.[1] The report has two main parts.
First, the cybersecurity research the government cited as the technical basis for the export control directive was Amazon's, not the government's. According to the report, Amazon's paper claims that, through a series of prompts, Fable 5 was made to serve up information that could be used in cyberattacks.[2] The paper is the same document Anthropic later said it had reviewed at the government's request and concluded was describing a "narrow, non-universal" capability that is "widely available from other models."[3]
Second, the conversations that converted that paper into a directive involved Amazon's CEO directly. According to the WSJ, Andy Jassy spoke with White House officials about Amazon's security concerns shortly before the export control letter landed at Anthropic at 5:21pm ET on Friday June 12.[1] The Verge, confirming the WSJ account, summarized the sequence: "Shortly after Jassy shared the company's findings with the government, it made the call to block [Fable 5's] use by foreign nationals."[2]
Amazon has not responded to a request for comment from The Verge.[2] The White House has not named the agency that sent the letter, has not cited a statute, and has not released Amazon's paper publicly. Anthropic said the letter "did not provide specific details of its national security concern" and that the company has only received "verbal evidence of a potential narrow, non-universal jailbreak."[3]
What Anthropic Said, and What It Did Not Say
Anthropic's June 12 statement did not name Amazon. It also did not name the agency that sent the letter, did not disclose the source of the underlying research, and did not describe a formal administrative process.[3] The company's public posture has been: comply with the directive, restore access as soon as possible, and dispute the technical finding on the record.
On the technical finding, Anthropic's statement is unambiguous. The company says it reviewed a report it believes is the basis of the directive and "validated that the level of capability displayed there is widely available from other models (including OpenAI's GPT-5.5), and is used every day by the defenders who keep systems safe."[3] The capability Anthropic describes is "essentially... asking the model to read a specific codebase and fix any software flaws." That is a code-review workflow that any contemporary model can perform, and that defenders use every day to find bugs before attackers do. Anthropic's own framing, the cleanest statement of the regulatory argument on either side, deserves to be quoted at length: "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."[3]
On the policy process, Anthropic is also on record with a more general position. The company has previously said "we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles."[3] The friction between those two statements, "yes, the government can do this, but not like this," is the policy fight the next 30 days will turn on. The WSJ story does not change the policy fight. It changes the audience for it.
What Anthropic has not said publicly is whether it has communicated directly with Amazon, AWS, or the AWS Bedrock team about the directive. Anthropic's largest cloud partner just used the US government to remove Anthropic's most advanced products from a market AWS Bedrock services. Anthropic's silence on that specific point is, at a minimum, a public-relations calculation. It is also a contractual one. AWS and Anthropic signed an expanded strategic collaboration in 2024, and Anthropic runs the bulk of its training and inference on AWS infrastructure. The directive does not change that relationship, but it puts the relationship on a different footing.
The "Jailbreak" That Wasn't One
The Verge's reporting introduces a critical second voice: Katie Moussouris, the founder and CEO of LutaSecurity, who has reviewed Amazon's paper and concluded that "It's not a jailbreak."[2] Moussouris is a long-standing figure in coordinated vulnerability disclosure. She co-authored Microsoft's first bug bounty program, helped stand up the Department of Defense's first bug bounty in 2016, and has been on the offensive-security side of the disclosure debate for twenty years. Her public statement is short, on the record, and cuts against the technical finding the directive rests on.
The former Commerce Department official Kate Koren, now a national-security analyst, told the WSJ that the White House's dislike of Anthropic may have influenced the decision.[2] The article cites the history. Anthropic and the Trump administration have been at odds for over a year. The friction has two specific public flashpoints. First, Anthropic's refusal to allow its AI to be used for mass surveillance of Americans or to power lethal autonomous weapons. Second, the February 2026 sequence in which the Trump administration instructed federal agencies to stop using Anthropic's AI and, hours later, Secretary of Defense Pete Hegseth designated the company a supply chain risk.[2] The Verge notes that "the government and the company seemed to have made amends" before this latest clash, and that the two had worked together to expand access to Mythos.[2] That expansion is now paused.
The 12grams of Carbon follow-up, written by a self-described AI doomer who has been running parallel agents on Fable 5 for months, captures the asymmetry more bluntly than any news write-up has. The author reads the directive against the historical record of "Friday evening" market-moving announcements from the Trump administration, the strike on Iran in February 2026, the Intel deal in August 2025, the 100% China tariff threat in October 2025, the Venezuela airspace closure in November 2025, and asks: "If any other government in the history of the United States took this step, there would be good" reasons named in the order itself, and a paper trail, and a public comment period, and a process.[5] The directive has none of those. The "narrow, non-universal" finding is the only stated basis, and the only person who has described the finding in print, Anthropic, says it is widely available from other models.
The Political Economy of the Anthropic Story
The story is no longer about whether Fable 5 is more dangerous than GPT-5.5. Anthropic's own statement, the Moussouris review, and the WSJ's account of Amazon's research all point in the same direction: the underlying capability is shared by every frontier model on the market. The story is about the political economy of how the most powerful AI deployments get gated, and about who has standing to trigger the gate.
Three structural facts make this story different from every prior "AI company asks government to slow a rival" story. First, the asking party is the largest cloud provider for the company that built the model. AWS hosts Anthropic's training and inference, runs the Bedrock service that retails Anthropic's models to AWS customers, and signed an expanded strategic collaboration with Anthropic in 2024 that has been the foundation of Anthropic's commercial growth. The party that benefits from the directive is, on paper, Anthropic's most important infrastructure partner.
Second, the request was made to the executive branch, not to a regulator, and the executive branch chose the export-control letter. The export-control route is the route with the least process. Anthropic has explicitly noted that the directive "does not adhere" to the principles the company believes a block on a frontier model should follow, "transparent, fair, clear, and grounded in technical facts."[3] The export-control letter has none of those attributes, by design. That is the policy hook the next 30 days will turn on.
Third, the request was made by the CEO personally, on the record with the Wall Street Journal, in a story that has now been read by every frontier-AI CEO in the country. OpenAI's Sam Altman, Google's Sundar Pichai, and Meta's Mark Zuckerberg now have a working playbook: identify the finding, share it with the White House, and let the export-control letter do the rest. The directive does not require a complaint from a competitor. The directive does not require a public rulemaking. The directive does not require the target company to be named. The directive does not require Amazon to disclose that it is the source of the technical research. The structure of the action is the precedent. The precedent will be available to every CEO who wants to use it.
The Anthropic Researcher Question
One under-reported fact in the WSJ's account is the human cost of the directive. The Verge's summary: "Complicating this issue is that many of Anthropic's researchers are foreign-born, meaning they were barred from accessing their own product."[2] Fable 5 and Mythos 5 are frontier AI models built by a research team that includes a substantial number of foreign-born researchers working on US visas. The directive, as written, requires that no foreign national inside or outside the United States, including foreign national Anthropic employees, be allowed to use the models.[3] The technical implementation Anthropic chose, disabling access for everyone, US citizen included, is the most cautious interpretation. The human implementation is that the people who built Fable 5 cannot run their own model.
The labor-market signal is the second-order story. Frontier AI talent is, in 2026, the most contested skilled labor pool in the US economy. The directive, by disqualifying any frontier-model company that uses foreign-born researchers from competing for the most advanced model deployments, is a back-door industrial policy. The companies that hire only US citizens are now structurally favored in any future export-control sweep. Whether the directive was intended that way is a separate question. Whether the directive operates that way is not.
What It Means for You Today
Three groups are affected, and the consequences diverge.
If you are a US-based individual subscriber or API customer. Nothing has changed since the day-1 disclosure. Fable 5 and Mythos 5 are not available to you. The 5.5 generation of Claude, Claude Opus 4.8, and the rest of the Anthropic model lineup are unaffected.[3] The WSJ story does not change the practical state of access. It changes the political context you should understand it in.
If you are a US enterprise AWS Bedrock customer who had enabled Fable 5 or Mythos 5. The directive and the WSJ story together are the most direct evidence yet that the AWS-Anthropic relationship is not a quiet partnership. The company that hosts Anthropic's models on its cloud is the same company that, through its CEO, triggered the export-control directive that pulled those models off the market. The 30-day data retention requirement we covered in the June 12 piece on Fable's hidden guardrails and the June 12 IT playbook for Mythos on Bedrock still applies to any future re-enable.[6][7] Expect Amazon's product, legal, and compliance teams to spend the next two weeks clarifying the Bedrock access policy in light of the directive. The clarifying language is the next data point worth watching.
If you are a frontier-AI competitor to Anthropic, or a frontier-AI investor, or a frontier-AI regulator. The new playbook is on the record. Find a research finding. Get the CEO in front of the White House. Wait for the export-control letter. The directive does not require a public rulemaking. The directive does not require the target to be named. The directive does not require the source of the technical research to be disclosed. The structure of the action is the precedent, and the precedent is now available to every CEO who wants to use it. The next letter may not be about a code review. The next letter may not come from Amazon. But the route the next letter will travel is the route Amazon opened in June 2026.
The Bottom Line
The Wall Street Journal reported on June 13, 2026 that Amazon CEO Andy Jassy's conversations with White House officials triggered the export control directive that forced Anthropic to disable Fable 5 and Mythos 5.[1] Anthropic's largest cloud partner used the executive branch to take Anthropic's most advanced products off the market. The technical finding the directive cites is a narrow code-review capability Anthropic says is "widely available from other models (including OpenAI's GPT-5.5)."[3] The cybersecurity researcher who has reviewed Amazon's paper says "It's not a jailbreak."[2] The executive-branch process the directive used is one Anthropic explicitly says "does not adhere" to the principles the company believes a frontier-model block should follow.[3]
The story is no longer about a jailbreak. It is about the new political economy of frontier AI. A competitor with the ear of the White House can now trigger a recall of a rival's most advanced products without a public rulemaking, without naming the target in advance, without disclosing the source of the technical research, and without any statutory process. The structure of the action is the precedent. The precedent will be available to every CEO who wants to use it.
Watch for three things over the next 96 hours. First, Amazon's first on-the-record response to the WSJ story. Second, Anthropic's first disclosure of the specific agency that sent the 5:21pm ET letter, the contract terms of the AWS Bedrock relationship as they apply to a recall, and the nationality-verification regime Anthropic intends to use to restore access. Third, the first statement from the White House AI policy staff on whether the export-control route will be the standing mechanism for any future "narrow, non-universal" finding on a frontier model. The answers to those three questions will determine whether the Jassy-to-Anthropic route is a one-off or a playbook.
Sources
- The Wall Street Journal: "Amazon CEO's Talks with U.S. Officials Triggered Crackdown on Anthropic Models" (June 13, 2026, paywalled, primary source for the Jassy conversations and the Amazon-origin cybersecurity research; the HN thread on the WSJ piece is item 48519092, 669 points at time of writing)
- The Verge: "Amazon security research reportedly led to the White House's Anthropic Fable ban" by Terrence O'Brien, June 13, 2026 at 9:39 PM UTC (open-access confirmation of the WSJ account, includes the Moussouris "It's not a jailbreak" quote and the Koren Commerce Department read)
- Anthropic: "Statement on the US government directive to suspend access to Fable 5 and Mythos 5" (June 12, 2026, 5:21pm ET directive, primary source for the technical dispute and the "narrow, non-universal jailbreak" framing)
- Hacker News: "Amazon CEO's talks with U.S. officials triggered crackdown on Anthropic models" thread (669 points, day-of WSJ coverage)
- theahura, "12 Grams of Carbon: There is a massive shadow hanging over this Fable thing" (June 13, 2026, follow-up analysis on the Friday 5:21pm timing pattern, the Trump administration's "Friday evening announcement" history, and the gap between the directive's stated process and the principles Anthropic has previously endorsed)
- State of Surveillance: "Anthropic Apologized for Fable's Hidden Guardrails. AWS Bedrock Customers Just Got the Real Story." (June 12, 2026, the silent-guardrails and AWS Bedrock data-residency context)
- State of Surveillance: "AWS Bedrock Mythos 30-Day Retention: The Enterprise IT Playbook" (June 12, 2026, the 30-day retention requirement and the data-boundary trade-off for Fable 5 and Mythos 5 on AWS Bedrock)