Tldr:
- The US Export Control Order imposed a global shutdown of Fable 5 and Mythos 5 models immediately for all users
- Anthropic says the restriction stems from alleged jailbreak concerns, but the calls originate in a narrow, irregular range.
- Claude’s other models remain available online due to the company’s compliance with the requirements of government national security directives
- The company disputes the seriousness of the claims, citing previous red team tests and the absence of verified malicious exploits
The US government’s export control directive was forced Anthropic To suspend global access to Fable 5 and Mythos 5. The order applies to all foreign nationals, including employees outside the United States.
The decision resulted in the immediate discontinuation of both models in all customer environments. According to internal communications, the directive was issued citing national security concerns related to potential vulnerabilities in the jailbreaking process.
The Export Control Order in Fable 5 and Mythos 5 shakes up access to human AI
AnthropicAI certain It received the directive at 5:21 PM ET from US authorities.
It takes an immediate hold to access Fable 5 and Mythos 5 worldwide. This restriction applies even to foreign employees working within the company.
The company stated that compliance was mandatory under export control rules.
As a result, it has disabled both models across its infrastructure. Other CLOUD models remain fully functional without restrictions.
The announcement highlighted that the directive affects users in all regions. Customers are out US Lost access at the same time with local users. The scope of the order reflects a broad national security classification.
Anthropic noted that the closure was sudden and not planned in advance.
Engineering teams implemented global deactivation procedures shortly after receiving the notification. The impact of the service outage on enterprise users and developers who rely on the models.
Fable 5 and Mythos 5 security concerns and jailbreak claims explained
The directive is said to stem from concerns about a potential jailbreak method targeting Fable 5.
Authorities believed this technique could detect Cybersecurity capabilities Under certain circumstances. The company reviewed the same offering internally.
Anthropic reported that the identified issue relates to narrow vulnerabilities already seen in other models. She added that similar vulnerabilities could be reproduced using publicly available artificial intelligence systems. The company did not specify evidence of a global jailbreak.
Internal evaluations showed that safeguards were extensively tested before launch.
The models have undergone thousands of hours of collaboration with external partners. This included government-linked AI safety institutes and external evaluators.
According to Anthropic, this report did not result in any verified malicious posting They are weak. The company said it had not received documentation of a large-scale exploit affecting typical safety systems. The case was described as limited in scope and unsystematic.
Although the severity of the directive was not approved, Anthropic complied with the legal requirements.
The company stressed ongoing discussions with regulators to restore access. It also reiterated that other models were not affected and continue to operate normally.
