Author: Xiong Lei

Editor: Xu Qingyang

On June 12 local time in the United States, a rare confrontation between government and enterprises regarding AI regulation suddenly ignited this Friday.

The U.S. government issued an export control directive to Anthropic, demanding an immediate suspension of all foreign entities' access to the AI models Fable 5 and Mythos 5. This directive has a broad scope, applying not only to foreign users outside the U.S. but also to foreign citizens within the U.S., including foreign employees within Anthropic.

As a compliance measure, Anthropic had to comprehensively shut down access to these two models for all users—this is currently the only feasible solution to ensure compliance, as the company cannot technically distinguish between “foreign entities” and “U.S. citizens” within the user base. Access to the company’s other models remains unaffected, and users will automatically revert to Claude Opus 4.8.

This emergency halt came unexpectedly. Fable 5 and Mythos 5 were officially released on June 9, only three days ago. The sudden removal of these two models has caused widespread shock in the tech world and AI community.

01 What are the two models?

To understand the core tension of this incident, one needs to understand what models Fable 5 and Mythos 5 are and why they have been under regulatory scrutiny from the beginning.

Mythos is a new model family developed by Anthropic that is higher in capability than the Opus series, representing the highest level of capability currently publicly deployable by Anthropic. The first Mythos model, Claude Mythos Preview, was released in April this year through the “Project Glasswing” initiative, with access strictly limited to a small number of partners due to its strong capabilities in cybersecurity, making it unsuitable for broad public access.

Fable 5 is the first publicly available Mythos-level model, exceeding all previously released models by Anthropic. It achieves top-tier results across nearly all benchmarks in software engineering, knowledge work, visual understanding, scientific research, etc.

To ensure public release, Anthropic equipped Fable 5 with specialized security mechanisms—automatically blocking responses in high-risk areas such as cybersecurity, biology, and chemistry, reverting to Claude Opus 4.8 for handling.

Mythos 5 is a version based on the same underlying model but equipped with fewer security protections, only available to institutions previously granted access through Project Glasswing, positioned as a professional tool for cyber defense practitioners and critical infrastructure operators. Both models are priced similarly, at $10 per million input tokens and $50 per million output tokens.

02 Trigger for the directive

Reports indicate that U.S. Secretary of Commerce Howard Lutnick sent a letter to Anthropic CEO Dario Amodei on June 12, announcing that Mythos 5 and Fable 5 would be subject to export controls. The immediate cause for this decision was a claim from another company that it could perform a "jailbreak" on Mythos, raising alarms about potential national security risks for the Trump administration.

It is reported that the Trump administration previously attempted to prevent Anthropic from releasing these two models but was unsuccessful—this subsequently prompted the government to adopt the more aggressive measure of export controls.

Faced with this sudden directive, Anthropic, while complying, issued a notably strong statement systematically rebutting the government's reasoning.

Anthropic argues that the "jailbreak" evidence held by the government only pertains to a very narrow and non-generalizable method of attack, essentially requiring the model to read specific code libraries and patch software vulnerabilities—this capability also exists in other publicly available models including OpenAI's GPT-5.5 and is used routinely by cybersecurity professionals for regular system maintenance.

In its statement, Anthropic clearly states that using the "presence of limited potential jailbreaking" as a standard for recalling deployed commercial models would effectively halt the rollout of all cutting-edge models across the entire industry. The company also emphasizes that the government's action did not follow the previously publicly called for transparent, fair, and technically factual legal procedures.

Anthropic stated that it would comply with the government’s legitimate directive while actively working to communicate with the government to restore access as soon as possible, and will announce more technical details within the next 24 hours. The company believes this incident stems from a misunderstanding and sincerely apologizes for the inconvenience caused to users.

Below is the full statement released by Anthropic titled "Statement on the U.S. Government's Suspension of Access to Fable 5 and Mythos 5":

The U.S. government, based on national security authorization, has issued an export control directive requiring the suspension of access to Fable 5 and Mythos 5 for all foreign citizens, regardless of whether they are located in the U.S. or abroad, including foreign employees of Anthropic. The practical effect of this directive is that we must immediately close access to Fable 5 and Mythos 5 for all users to ensure compliance. Other Anthropic models remain unaffected.

We received this directive today at 5:21 PM Eastern Time. The letter did not specify the exact national security concerns. To our understanding, the government believes it has acquired a method to bypass Fable 5's security protections, known as "jailbreaking." We reviewed a demonstration of this technology and found it can only identify a small number of previously known, low-impact vulnerabilities. These vulnerabilities are relatively simple, and other publicly available models can also detect the same issues without the need for jailbreaking.

Anthropic outlined our position on its security measures in a blog post about the release of Fable, as follows:

We have established robust security measures that significantly reduce the likelihood of Fable being misused for cybersecurity and related tasks. In fact, some users have reported that our protective measures are overly stringent.

Weeks before the release of Fable, Anthropic collaborated with the U.S. government, the UK's AI safety research institute, several third-party organizations, and internal teams to conduct thousands of hours of red team testing on Fable's security mechanisms. The results indicated that Fable's security performance was significantly superior to that of any previously deployed models.

So far, no testers have been able to find a "generic jailbreak" method—that is, a method capable of entirely bypassing the model's security protections to unlock broad capabilities for cyber attacks.

We believe that no model provider can achieve perfect jailbreak protection at this time. All protective mechanisms in the industry are susceptible to being breached by “non-general jailbreaks” (i.e., obtaining partial network information under specific conditions), and a generic jailbreak method could emerge in the future. This is something we clarified at the time of the Fable 5 release.

Given that perfect jailbreak protection is currently not feasible, Anthropic has implemented a layered defense strategy for Fable 5. Our goal is to make jailbreaking attacks either very narrow in scope (for non-general jailbreaks) or prohibitively costly (for generic jailbreaks), supplemented by comprehensive monitoring to quickly detect and respond to any successful attacks. This is also why Anthropic requires client data to be retained for 30 days—although this policy has had a practical impact on our client relationships, it helps us study and respond to jailbreak risks.

We stand by this layered defense strategy. It has effectively reduced the risks associated with Fable to a level comparable to that of existing industry-deployed models.

So far, we have not even received any formal disclosures regarding potential non-general jailbreaks that could lead to harmful consequences. The potential jailbreak cases disclosed to us have either been completely harmless or merely represent minor findings that do not showcase Mythos' exclusive capabilities.

Currently, the U.S. government has only provided us with verbal evidence regarding a potential, limited non-general jailbreak method, which essentially requires the model to read specific code libraries and patch software vulnerabilities. To our knowledge, this potential jailbreak method has been shared with the government. We reviewed a report—we believe this report forms the basis for the government's issuance of the directive—and verified that the capability levels demonstrated in the report are also prevalent in other models (including OpenAI's GPT-5.5), and are routinely used by cybersecurity professionals to maintain system safety. We will publish more details within the next 24 hours.

We will comply with the government's legal directive to close access to Fable 5 and Mythos 5 for all users. However, we do not agree that finding a limited potential jailbreak method is sufficient reason to recall a commercial model that has been deployed to hundreds of millions of users. If this standard were applied to the entire industry, it would effectively stop the deployment of new models by all leading model providers.

We have publicly stated that the government should have the authority to prevent the deployment of unsafe models according to transparent, fair, clear, and fact-based legal procedures. This action did not adhere to the aforementioned principles.

We sincerely apologize for the inconvenience caused to users. We believe this incident is due to a misunderstanding, and we are actively working to restore access as soon as possible.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。