The “most capable” AI model in the Anthropic Cloud Mythos leaks, and is considered a major cybersecurity threat

Claude creator Anthropic is developing a new AI model called Claude Mythos, which has been described internally as the company’s most capable model yet, with draft materials about the system leaked online this week.

The existence of the model was first I mentioned by luck on Thursday after unpublished files linked to the Anthropic blog were discovered in a publicly accessible data cache. A spokesperson for Anthropy confirmed the availability of the model for publication.

An Anthropic spokesperson said: “We are developing a general-purpose model with tangible advances in reasoning, cryptography and cybersecurity.” luck. “Given the strength of its capabilities, we are looking at how to release it. As is standard practice across the industry, we are working with a small group of early access customers to test the model. We view this model as a step change and it is the most capable model we have built to date.”

In archived development page Reviewed by DecryptionAnthropic described Mythos as “the most powerful AI model we have ever developed.”

“Mythos is a new name for a new level of models: bigger and smarter than our Opus models — which were, by far, our most powerful,” Anthropic wrote. “We chose the name to evoke the deep connective tissue that binds knowledge and ideas together.”

According to Anthropic, the Mythos scored “significantly higher” than the Claude Opus 4.6 on tests of software coding, academic reasoning, and cybersecurity.

The Mythos leak appears to have originated from draft material stored in an insecure content management system. according to luck,A human restricted public access to the data store after being notified that the files were searchable online. The company attributed the exposure to human error in configuring its content management system (CMS) tools.

However, Anthropic’s documentation described the Mythos as the first version of the new model, and described the second version internally as the “Capybara”, which the company also placed over the existing top-tier Opus models.

The draft materials also highlighted concerns about the system’s potential impacts on cybersecurity.

“Although Mythos is currently far ahead of any other AI model in cyber capabilities, it heralds a coming wave of models that can exploit vulnerabilities in ways that far exceed the efforts of defenders,” the company wrote.

Because of these risks, the company said it plans to release the model cautiously, starting with a limited early access rollout targeting organizations working on cybersecurity defense.

Anthropic did not respond immediately Decryption Request for comment.

While Anthropic deleted the blog post, news of the leak quickly spread to the financial markets.

Shares of several cybersecurity companies decreased After reports surfaced, including Palo Alto Networks (Bano), which was down about 7%, and CrowdStrike (Raw) which decreased by about 6.4%. Meanwhile, Zscaler (ZS) and decreased by about 5.8%, and Fortinet (Fascinated) fell about 4% during Friday trading, according to Yahoo Finance.

The selling reaction reflects a similar market response to the unveiling of a new human product. In February Anthropy unveil Claude Cowork, an artificial intelligence system designed to automate complex tasks in the workplace — including contract review and compliance — has led to widespread sales across software and professional services companies.

This sell-off wiped out nearly $285 billion in market value as investors reassessed the long-term impact of AI agents on enterprise software companies.

“The market response was a signal, not that AI agents would immediately replace these companies, but that investors took into account the structural risks that core model providers could now compete directly with the software layer,” said Scott Dillane, founder of Nexatech Ventures. Decryption at that time. “That’s a polite way of saying that if Anthropic can create a tool for legal workflow within a company, what’s to stop it from doing the same for finance, procurement, or HR?”