Author: Claude, Deep Tide TechFlow
Deep Tide Introduction: Citing national security concerns, the US government on June 12th suddenly ordered Anthropic to shut down its two strongest AI models, Fable 5 and Mythos 5. Impacted by this, the Hyperliquid perpetual contract tracking its expected IPO valuation fell by 3.7% to around $1,627 on Saturday, significantly retreating from the post-model-launch high of over $1,800. Anthropic publicly expressed strong opposition, stating that the government demanded a recall based solely on an oral briefing of a "narrow jailbreak vulnerability," and warned that applying such a standard across the entire industry "would essentially halt all new model deployments from frontier model providers."
A company founded on "AI safety" is now forced to shut down its strongest products by a government safety order.
According to reports from CoinDesk and other media outlets, on June 12th (Friday) afternoon, the US government issued an export control directive to Anthropic, requiring it to suspend all access to Claude Fable 5 and Claude Mythos 5 for foreign nationals—whether inside or outside the US, including the company's own foreign employees. Unable to implement selective restrictions without affecting compliance, Anthropic simply disabled these two models for all users globally; other Claude models remain unaffected. In a statement, Anthropic clearly stated its disagreement with the government's rationale, calling it "a misunderstanding," and is working to restore access as soon as possible.
Directive Cites National Security, One Letter Triggers Full Shutdown
According to The Wall Street Journal (as cited by Quartz), US Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei outlining the relevant restrictions. Anthropic revealed that the company received the directive at 5:21 PM (Eastern Time) that day, but the letter did not provide specific details of the national security concerns.
Anthropic stated that, to its understanding, the government believes it has discovered a method to bypass or "jailbreak" Fable 5. Company executives explained that the cited vulnerability involves a narrow prompt engineering technique used to make the model review pre-existing, minor software bugs in code, emphasizing that similar outputs can be easily obtained from competitor models like OpenAI's without any special evasion tactics.
Anthropic's wording in its statement was unusually direct: "We are complying with a legal directive from the government and removing access to Fable 5 and Mythos 5 for all users. But we disagree that discovering a narrow potential jailbreak vulnerability should constitute grounds for recalling a commercial model already deployed to hundreds of millions of people." The company warned that if this standard were applied across the industry, it "would essentially halt all new model deployments from frontier model providers."
The company also noted that, as of now, the government has only provided oral evidence—no written documentation, technical report, or formal disclosure through established channels—and promised to publish further technical analysis within 24 hours to illustrate the limited nature of the issue.
Quasi-IPO Market Reacts Instantly, Hyperliquid Perpetual Contract Plunges
The reason this regulatory news became a financial event is that Anthropic is not yet publicly listed, and the crypto market has already built a "shadow pricing" venue for it.
According to CoinDesk, the Anthropic perpetual contract on Hyperliquid fell about 3.7% to around $1,627 on Saturday. In the days following the Fable 5 launch, the contract had been trading near its all-time highs above $1,800. The contract's open interest is approximately $8.6 million. While this volume is small relative to larger proxy instruments for the private market, it represents a notable signal for a company that has yet to file for an IPO.
The contract's price roughly corresponds to an implied valuation for Anthropic in the trillions of dollars—a price of $1,638 equals an implied valuation of about $1.638 trillion. Within 24 hours of the news breaking, the contract's 24-hour trading volume was approximately $190,000, with a funding rate of 0.0056%.
The impact was not limited to the contract market. According to Bitcoin.com, the ANTHROPIC token issued by Prestocks on the Solana blockchain—which claims to provide spot-like exposure through a Special Purpose Vehicle (SPV) structure—fell 9.11% in 24 hours to $655.32, and the total value locked in the related liquidity pool shrank 18.09% to about $126,000. It should be noted that Anthropic has previously publicly warned that any unauthorized tokenization or derivatives claiming to offer exposure to its equity may be considered invalid, and the aforementioned products are not endorsed by or affiliated with Anthropic.
Shut Down Just Days After Launch, Fable 5 Was a Milestone
The two halted models were released just days ago.
According to multiple reports, Fable 5 and Mythos 5 were launched around June 9th and were seen as a significant leap forward in capabilities for software engineering, scientific reasoning, vision tasks, and long-context handling. Fable 5 marked Anthropic's first public release of a model at the Mythos level—a tier positioned above the original flagship Opus. The two models are built on the same underlying architecture, primarily differing in output control: Fable 5 has a built-in classifier for intercepting responses in high-risk areas like cybersecurity, while Mythos 5 is available to a separately vetted, trusted user group.
As summarized by MarkTechPost, Fable 5's classifier is a separate AI system designed to identify potential misuse. When queries touch on areas like cybersecurity, biochemistry, or model distillation, processing falls back to Claude Opus 4.8, with users notified of the fallback each time; the trigger rate is less than 5%. Notably, Mythos 5 was also supporting critical cybersecurity defense work through projects like Project Glasswing—raising the question of whether the temporary disablement might actually hinder infrastructure protection.
The Safety Narrative Backfires, Adding Uncertainty to IPO Process
For Anthropic, the awkwardness of this incident lies in its timing and logic.
TechCrunch reported the story under the headline "Anthropic’s safety warnings may have just backfired"—it's a bitter pill to swallow for a company that has built its entire corporate identity on "responsible AI development" to now be ordered to shut down its strongest models based solely on an oral briefing. And the company has publicly voiced this discontent.
This controversy coincides with a sensitive period as Anthropic advances its IPO preparations. The company recently confidentially submitted registration documents, joining the ranks of major AI players planning public listings. Although tools like Hyperliquid have relatively thin liquidity, the contract's decline reflects investor caution regarding its near-term valuation and listing path. Such sudden regulatory actions add new variables to an already uncertain IPO process.
Anthropic stated it is actively communicating to resolve the matter and restore service as soon as possible, viewing the directive as "based on a misunderstanding." The ultimate outcome of this incident may influence the future dynamics between regulators and innovators, and even shape how export rules and safety assessments apply to evolving AI technology.
