Anthropic’s pre-IPO shares have decreased as the US government orders the suspension of its most advanced AI models. Following a jailbreak report, Washington mandated the removal of Fable 5 and Mythos 5 from all users. While Anthropic complied, they strongly criticized the decision, claiming it sets a precedent that could impede the entire AI sector.

— The US government directed Anthropic to cease access to its two leading AI models, Fable 5 and Mythos 5, due to national security concerns over a potential jailbreak.

— Anthropic has contested the order, asserting that the government has only presented verbal evidence of a limited, previously recognized vulnerability that can also be detected by similar models, including OpenAI’s GPT-5.5.

— The market reacted unfavorably to the shutdown, with the Anthropic perpetual contract on Hyperliquid dropping approximately 3.7 percent, raising further concerns about the company’s pre-IPO valuation and future listing.

On Friday, the US government mandated Anthropic to disable its two most capable AI models, leading to an immediate sell-off in the pre-IPO market for Anthropic shares.

The Anthropic perpetual on Hyperliquid, a cash-settled contract reflecting traders’ expectations for the company’s equity pricing, fell 3.7% on Saturday to roughly $1,627, according to available data.

The contract had previously been trading close to record highs above $1,800 shortly after the launch of Fable 5. Open interest on the contract is around $8.6 million, relatively small compared to the SpaceX perpetual but significant for a company yet to file for an IPO.

Anthropic reported receiving the directive from the US government at 5:21 p.m. ET on Friday. The order, citing national security concerns, mandated the company to suspend all access to Fable 5 and Mythos 5 for any foreign national, regardless of their location. Subsequently, a complete shutdown was executed for all users.

The government informed Anthropic that it had discovered a method to bypass, or jailbreak, Fable 5. Anthropic assessed the technique and stated that it was limited, not a universal jailbreak, and involved recognizing a few previously known, minor vulnerabilities. They indicated that other publicly available models, such as OpenAI’s GPT-5.5, could identify the same vulnerabilities without any bypass.

The company noted that the government has so far offered only verbal evidence of a possible narrow jailbreak, which they described as essentially asking the model to analyze a codebase and rectify software errors, a routine task for defenders.

They argued that applying this standard industry-wide «would effectively halt all new model deployments for all frontier model providers.»

Anthropic has centered its entire brand around the development of safety-first AI and is now publicly contesting a national security directive, claiming the government’s evidence does not meet its own stated criteria.

The company is expected to provide further details regarding the specific jailbreak within 24 hours.

The cryptocurrency market is now factoring in the shutdown as a negative development for the IPO scenario, and the decline of the Anthropic perpetual from its post-launch peaks reflects this sentiment. The primary question for the company’s public listing ambitions will be whether the government’s order is reversed, modified, or expanded to include other model categories once Anthropic submits its technical response.