An accidental leak has now been officially confirmed by AI company Anthropic regarding its most powerful AI model yet.
The model, now known as “Claude Mythos,” was originally uncovered in a report from Fortune. Anthropic has since confirmed the details about the leak to the outlet.
The data leak included details about the upcoming release of the Claude Mythos AI model, an “exclusive CEO event,” and other internal assets such as PDFs and images, according to Fortune. In total, the leak included nearly 3,000 assets that Anthropic had not previously published.
Claude Mythos leak details
According to Anthropic, the leak was the result of an issue with the content management system (CMS) that the company uses. The data was uploaded to the CMS by the company. However, Anthropic failed to mark the items as private so the data was stored in a publicly accessible data lake.
Some of the data was unused assets from past announcements from Anthropic. Other leaked data was internal corporate information such as employee details or info about an invite-only CEO event.
Also found within the leaked data, however, was information about Claude Mythos in an unpublished blog post draft.
Mashable Light Speed
Claude Mythos and Capybara
According to the leaked draft about Claude Mythos, Anthropic has completed its training of the model and refers to it as “by far the most powerful AI model we’ve ever developed.”
In a statement provided to Fortune after the leak, an Anthropic spokesperson described Claude Mythos as an AI performance “step change” and said the model is “the most capable we’ve built to date.” Anthropic says that Claude Mythos is currently in the trial stage and available to select “early access customers.”
In addition to the new AI model, the leaked post also mentions a new, unreleased AI model tier from Anthropic called Capybara.
Anthropic currently offers three different tier levels for each of its AI models: Opus, Sonnet, and Haiku. Opus is currently its most powerful and most expensive tier whereas Haiku is the smallest and least expensive. Capybara would sit even above Opus as Anthropic’s largest and most intelligent AI model tier.
Capybara cybersecurity risks
The leaked post also details Anthropic’s concerns about the cybersecurity risks that Claude Mythos brings to the table, too. It appears that Anthropic believes that Claude Mythos could be used by hackers to run cyberattacks.
“In preparing to release Claude Capybara, we want to act with extra caution and understand the risks it poses — even beyond what we learn in our own testing,” reads the leaked Anthropic post. “In particular, we want to understand the model’s potential near-term risks in the realm of cybersecurity — and share the results to help cyber defenders prepare.”
Anthropic described its unreleased model as “currently far ahead of any other AI model in cyber capabilities” and that the model “presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.”
The company said in the leaked draft blog that it was providing early access to organization to give them “a head start in improving the robustness of their codebases against the impending wave of AI-driven exploits.”
