Anthropic's new artificial intelligence model named Claude Mythos, although not officially launched, has quickly become the focus of attention of the technology world.
Leaked information shows that this could be a major step forward in the race to develop large-scale language models (LLM), while raising many concerns about cybersecurity risks.
According to leaked internal documents, Claude Mythos is described as the most likely model Anthropic has ever built, surpassing even the previous version, Claude Opus 4. 6.
This information appeared after security researchers discovered an unpublished blog draft placed in a public data warehouse due to a system configuration error.
Anthropic later confirmed the incident, saying the cause came from "human error". However, the leaked content also revealed many important details about the company's AI development orientation.
Claude Mythos is said to belong to the new product line called Capybara, which is a model generation designed more powerfully than the previous Opus line.
While Anthropic currently maintains three model levels including Opus, Sonnet and Haiku, the Capybara line is expected to set new performance standards, although operating costs are significantly higher.
The most notable point lies in the capacity related to cybersecurity. According to leaked information, Mythos achieved outstanding results in tests on programming, academic thinking and especially the ability to detect and exploit security vulnerabilities. This made Anthropic itself admit potential risks.
The company warns that the model may be exploited by bad actors to carry out large-scale cyber attacks with a higher level of sophistication than before.
Even, Mythos is said to far surpass other AI models in terms of network security capabilities, posing new challenges for defense systems.
Therefore, Anthropic's deployment strategy is considered cautious. Instead of widespread release, the company currently only provides early access to a small group of users, mainly organizations and security experts.
The goal is to help them prepare for the wave of cyberattacks that may be supported by AI in the near future.
The appearance of Claude Mythos is not only a technological advance, but also a warning about a new era of cybersecurity, where AI is both a protective tool and can become a threat if not tightly controlled.
Currently, the official launch date of this model has not been announced.
