TL;DR
White House AI adviser David Sacks says Anthropic refused to fix a serious jailbreak in Fable 5, leading to U.S. export controls that pushed Anthropic’s top models offline. Anthropic says the flaw was narrow, already known and not enough to justify the action. The core evidence remains non-public, leaving users, developers and policymakers to weigh competing claims.
White House AI adviser David Sacks said over the weekend that Anthropic’s Fable 5 model was restricted because the company refused to fix a jailbreak that, in his account, could restore Mythos-class cyber capabilities; Anthropic disputes that account, saying the government provided no specific technical detail and that the reported flaw was narrow. The clash matters because Anthropic’s most powerful models were pulled offline under U.S. export controls on evidence that remains largely non-public.
Sacks, who is co-chair of the President’s Council of Advisors on Science and Technology, said a "highly credible trusted partner" found a way around Fable’s safeguards. According to his account, the administration asked Anthropic CEO Dario Amodei to fix the issue or pull the model, and export controls followed after he refused.
Anthropic has given a sharply different version. The company said the government did not provide specific technical details, described the demonstration as involving minor and already-known flaws, and argued that comparable public models could produce similar outputs without the same kind of bypass.
Reporting cited by Thorsten Meyer AI, including accounts attributed to Semafor and other outlets, has pointed to Amazon as a possible source of the warning. Amazon has not confirmed the specifics publicly. That matters because Amazon is simultaneously an Anthropic investor, a major cloud provider for the company, and a competitor in AI models.
The Safety Card, Played From Every Side
● ContestedA White House adviser says Anthropic refused to fix a cyberweapon jailbreak and got banned for it. Anthropic says the flaw is trivial. Almost every fact that would settle it is non-public — and “safety” is now the card every side is playing.
Both are claims, not findings. They don’t disagree on tone — they disagree on what the bypass actually is.
- A “highly credible trusted partner” found a jailbreak of Fable’s guardrails.
- The admin asked Amodei to fix it or pull the model. He refused.
- So the export control was issued — “reluctantly.”
- It restores operability of a cyberweapon; calling that “not serious” is indefensible.
- The government gave no specific technical detail.
- The demo found a few minor, already-known flaws.
- Other public models (incl. GPT-5.5) do the same without a bypass.
- A “narrow potential jailbreak” shouldn’t recall a model used by hundreds of millions.
Per reporting by Semafor (carried by Fortune and others), the entity that flagged the jailbreak was Amazon — with CEO Andy Jassy reportedly in contact with the administration. Amazon hasn’t confirmed specifics. Flagging a real risk is what a good partner does — but Amazon wears three hats at once, and none of them is neutral.
Each actor’s safety claim points toward its own advantage.
The entire evidentiary record is a matter of trusting parties who each have a reason to shade it.
A transparent, technically grounded, independently reviewable process — which is, notably, exactly what Anthropic says it wants, and exactly what would also constrain Anthropic. The reason to demand it isn’t loyalty to anyone; it’s that the alternative is decisions made on secret evidence and adjudicated in dueling press statements.
Independent commentary, produced with AI assistance under human editorial oversight; the views are the author’s own and may change. This is analysis and opinion, not investment, financial, legal, or technical advice, and it concerns an actively developing situation in which key facts are disputed and non-public. Claims attributed to David Sacks reflect his June 13, 2026 statement on X; claims attributed to Anthropic reflect its published statements; reporting on Amazon’s role reflects accounts published by Semafor and others — all read as of June 15, 2026, and presented as the claims of those parties, not as established fact. Characterizations are the author’s interpretation, offered in good faith and open to rebuttal. References to specific people, companies, and government actions are factual and analytical, not partisan, and imply no affiliation or endorsement.
Secret Evidence Tests AI Oversight
The dispute is now about more than one model release. It is a test of whether frontier AI restrictions can be imposed on the basis of classified or otherwise non-public evidence while still giving affected companies, customers and foreign partners enough information to judge the decision.
The case also shows how safety claims can serve different interests at once. The government is invoking safety to justify forceful controls on commercial AI. Anthropic, a company long associated with stronger AI safety rules, is now arguing that the government’s action overstates the danger. If Amazon played the role described in reporting, a safety warning may also have affected a company with which it has financial and competitive ties.
Privacy Tools in the Age of AI: Practical Strategies with VPNs, Secure DNS, Private Relay and Intelligent Defenses (Build Your Own VPN)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Fable 5 Followed Mythos Concerns
Sacks’ account rests on a specific claim about the relationship between Fable and Mythos. He said Fable is effectively Mythos with guardrails, meaning that a successful bypass could expose capabilities Anthropic itself had treated as too sensitive for broad release.
Anthropic’s response focuses on severity and process. The company has not accepted the government’s framing that the reported issue amounts to a serious cyberweapon-enabling failure. It says the flaw was limited and that the administration did not provide the technical evidence needed to verify or repair the alleged problem in the way officials demanded.
“A "highly credible trusted partner" found a jailbreak of Fable’s guardrails.”
— David Sacks, White House AI adviser, via X
It is not yet clear what exact method was used to bypass Fable’s safeguards, whether the result restored the cyber capabilities Sacks described, or whether the same output could be reproduced across other leading models without a special bypass. No public methodology, technical report, CVE-style record or independent review has been released.
The identity and role of the "trusted partner" also remain unresolved in public. Reports have pointed to Amazon, but Amazon has not publicly confirmed the specifics described in those accounts. Without the underlying evidence, readers are left comparing claims from parties with strong institutional interests.
The Ban Timeline Becomes The Test
The next signal will be whether Anthropic can satisfy U.S. officials quickly and restore access after changes to the models or their safeguards. A rapid return would suggest the dispute was fixable through a targeted patch or policy adjustment; a longer standoff would point to a deeper disagreement over standards, evidence and control of frontier AI systems.
For now, the strongest demand from outside observers is procedural: publish enough methodology for independent review, or create a trusted review process that can test the government’s claims without exposing sensitive details.
Key Questions
What happened to Anthropic’s Fable 5 and Mythos 5 models?
Anthropic took its top models offline after U.S. export controls targeted access to the systems. The government linked the action to national security concerns, while Anthropic disputes the severity and handling of the issue.
What did David Sacks allege?
Sacks said a trusted partner found a jailbreak in Fable’s safeguards and that Anthropic refused to fix or pull the model when asked. He framed the issue as a risk that could expose Mythos-class cyber capabilities.
What is Anthropic’s response?
Anthropic says the government did not provide specific technical detail and that the reported flaw was narrow, minor and reproducible in other public models. The company argues the response was too broad for the issue described.
Was Amazon involved?
Semafor and other reporting cited in the source material have identified Amazon as a possible source of the warning. Amazon has not publicly confirmed the specific role described in those reports.
Is the jailbreak confirmed?
A jailbreak has been alleged by Sacks and disputed by Anthropic, but the technical evidence has not been released publicly. That leaves the severity of the issue unresolved.
Source: Thorsten Meyer AI
