Il y a quelques jours, Anthropic annonçait Claude Mythos 5 et Claude Fable 5.
Claude Mythos 5 est la version sans restriction de cybersécurité et biologie. Il n’est accessible qu’à un cercle restreint d’organisations sélectionnées dans le cadre du Project Glasswing
De son côté, Claude Fable 5, le premier modèle de la classe Mythos accessible au grand public. Claude Fable 5 se comporte comme Mythos mais lorsque des requêtes un peu trop sensibles sont détectées, le système bascule vers Opus 4.8.
A l’occasion de ces annonces, Anthropic expliquait que 80 % du code était désormais effectué par l’IA et que les agents automatisaient même cette prise en charge. On comprend la raison de l’accélération que l’on observe actuellement.
Mais ces systèmes sont considérés si puissants, que le gouvernement américain vient d’en interdire l’usage à tout non américain, qu’il soit ou non présent sur le territoire des États-Unis. Cette décision démontre, si besoin était que la notion de dépendance technologique est bien une réalité. Elle avait été expérimentée sur le juge de la CPI Nicolas Guillou et certains de ses collègues. Un cap a été franchi avec une mesure de portée plus générale.
Anthropic s’est fendu d’un communiqué (dicté en partie par le gouvernement américain) expliquant que « le gouvernement américain, invoquant ses prérogatives en matière de sécurité nationale, a émis une directive de contrôle des exportations suspendant tout accès à Fable 5 et Mythos 5 pour tout ressortissant étranger, qu’il se trouve à l’intérieur ou à l’extérieur des États-Unis, y compris les employés d’Anthropic de nationalité étrangère ».
On imagine assez bien l’imbroglio qui va être créé si des développeurs étrangers travaillent sur ces modèles.
« Nous avons reçu cette directive du gouvernement aujourd’hui à 17h21 (heure de l’Est). La lettre ne précise pas les motifs de sécurité nationale invoqués », poursuit le communiqué.
Mais il conclut en indiquant que « comme nous l’avons déclaré publiquement, nous pensons que le gouvernement doit disposer de la capacité de bloquer des déploiements dangereux, dans le cadre d’un processus législatif transparent, équitable, clair et fondé sur des faits techniques. Cette action ne satisfait pas à ces principes.
Nous présentons nos excuses pour cette interruption de service à nos clients. Nous considérons qu’il s’agit d’un malentendu et nous travaillons à rétablir l’accès dans les meilleurs délais ».
Statement on the US government directive to suspend access to Fable 5 and Mythos 5
12 juin 2026
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Anthropic models will not be affected.
We received the directive from the government today at 5:21pm (ET). The letter did not provide specific details of its national security concern. Our understanding is that the government believes it has become aware of a method of bypassing, or “jailbreaking” Fable 5. We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.
Anthropic’s posture with respect to Fable’s safeguards, as laid out in our launch blog post, is the following:
- We have instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity (among others). In fact, our safeguards are so strong that many users have complained that they are overly broad.
- In the weeks leading up to the launch of Fable, Anthropic worked with the US government, the UK AISI, multiple private third-party organizations and internal teams to red-team Fable’s safeguards for thousands of hours in total.
- These tests showed that Fable’s safeguards are substantially more effective than those of any previously deployed model.
- No testers have yet been able to find a universal jailbreak—a jailbreak method that can very broadly bypass the model’s safeguards, unblocking a wide range of cyber capabilities.
- We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks (which can elicit some cyber information in specific circumstances), and it is likely that universal jailbreaks will eventually be found in the future. We stated this clearly when we released Fable 5.
- Given that perfect jailbreak resistance does not appear to be possible today, Anthropic adopted a defense in depth strategy with Fable 5. We aimed to make jailbreaks either narrow (in the case of non-universal jailbreaks) or very expensive to produce (in the case of universal jailbreaks), and to combine this with thorough monitoring to quickly detect and shut down any successful attacks. This is also why Anthropic has required 30-day retention of customer data with Fable—a policy change that carries real costs for us with customers, but that allows us to research and mitigate jailbreaks.
- We stand by this defense in depth strategy. It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed across the industry.
- We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result. The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift.
To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws. Our understanding is that one potential jailbreak was shared with the government. We have reviewed a report that we believe is the basis of the government’s 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. We will share more details over the next 24 hours.
We are complying with the government’s legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.
As we have stated publicly, 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.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.