Anthropic is building a new artificial intelligence model called Claude Mythos, described internally as the most powerful system the company has ever developed. The model’s existence came to light this week after unpublished files tied to Anthropic’s blog were found in a publicly accessible data cache. The story was first reported by Fortune on Thursday, and an Anthropic spokesperson subsequently confirmed the model’s existence to the publication.
In a statement to Fortune, an Anthropic spokesperson said the company is developing a general-purpose model with meaningful advances in reasoning, coding, and cybersecurity. The spokesperson described Mythos as a step change and the most capable model Anthropic has built to date. The company added that it is working with a small group of early-access customers to test the system, citing the strength of its capabilities as the reason for a deliberate release approach.
An archived development page reviewed by Decrypt quoted Anthropic describing Mythos as representing a new tier of model, larger and more intelligent than its existing Opus models, which had previously been the company’s most powerful. Anthropic stated it chose the name Mythos to evoke the deep connective tissues that link together knowledge and ideas. According to the documents, Mythos scored dramatically higher than Claude Opus 4.6 on tests covering software coding, academic reasoning, and cybersecurity.
The leaked materials appear to have originated from draft content stored in an unsecured content management system. Anthropic restricted public access to the data store after being notified that the files were searchable online, attributing the exposure to human error in the configuration of its CMS tools. The documents also referenced a second version of the model, referred to internally as Capybara, which the company similarly positioned above its current top-tier Opus models.
The draft materials flagged significant concerns about Mythos’s cybersecurity implications. Anthropic wrote that although Mythos is currently far ahead of any other AI model in cyber capabilities, it signals an upcoming wave of models that can exploit vulnerabilities in ways that outpace the efforts of defenders. In response to those risks, the company said it plans to begin with a limited early-access rollout aimed at organizations working on cybersecurity defense.
News of the leak quickly moved financial markets. Shares of several cybersecurity firms declined during Friday trading, according to Yahoo Finance. Palo Alto Networks fell approximately 7%, while CrowdStrike dropped roughly 6.4%. Zscaler declined around 5.8% and Fortinet slipped about 4%.
The market reaction echoes a similar episode in February, when Anthropic unveiled Claude Cowork, an AI system designed to automate complex workplace tasks including contract review and compliance. That announcement triggered a broad sell-off across software and professional-services companies, erasing roughly $285 billion in market value as investors reassessed the long-term impact of AI agents on enterprise software businesses. Nexatech Ventures founder Scott Dylan told Decrypt at the time that the market’s response reflected investors finally pricing in the structural risk that foundation model providers can now compete directly with the software layer. Anthropic did not immediately respond to Decrypt’s request for comment on the Mythos leak.
Originally reported by Decrypt.
