Anthropic's New AI Model with 'Unprecedented Cybersecurity Risks'? The Most Ironic Leak

As companies continue to invest heavily in AI model development, the race for technological supremacy shows no signs of slowing down. Anthropic is working on a new AI model called 'Claude Mythos'. According to a company spokesperson, this model represents a 'step change' in AI proficiencies and is the 'most capable we've built to date'. It is described as a general-purpose model with meaningful advances in reasoning, coding, and cybersecurity.

The news of this development comes at a crucial time, where companies are still investing large sums of money in the development of AI models, and the race for technological supremacy shows no signs of slowing down. This new model could mark a milestone in the evolution of AI, but it also raises questions about its impact on digital security.

The irony of the situation is notable: a data leak, caused by a major security lapse in Anthropic's content management system, revealed details about 'Claude Mythos'. A draft blog obtained by Fortune, which was 'available in an unsecured and publicly searchable data store,' claimed that the new model 'poses unprecedented cybersecurity risks'.

This incident puts the company to the test, which has received a lot of media attention for its Claude Code and Claude Cowork tools. The leak also revealed a 'new tier' of AI models, called Capybara, of which Mythos seems to be a part, but it is not yet clear how Capybara fits in with Anthropic's existing tiers: Opus, Sonnet, and Haiku.

According to the leak, 'compared to our previous best model, Claude Opus 4.6, Capybara gets dramatically higher scores on tests of software coding, academic reasoning, and cybersecurity, among others'. However, the company warns that while the model could score higher in cybersecurity tests, it could represent a major challenge for existing cybersecurity defenses.

Anthropic has expressed its intention to act with caution and understand the risks posed by the model, even beyond its own testing. The company seeks to understand the model's near-term risks in the realm of cybersecurity and share the results to help cyber defenders prepare.

The risks appear to be real enough for cybersecurity stocks to plunge on Friday following the latest news. Anthropic previously admitted that hackers used its Claude AI model to automate cybercrimes targeting banks and governments. According to the company's November blog post, a Chinese state-sponsored group exploited the AI's agentic capabilities to infiltrate 'roughly thirty global targets and succeeded in a small number of cases' by 'pretending to work for legitimate security-testing organizations' to sidestep Anthropic's AI guardrails.

This history of incidents raises serious questions about the security of AI models and their potential to be used for malicious purposes.

The announcement of a 'revolutionary' AI model is common in the industry, but it remains to be seen whether 'Claude Mythos' will live up to expectations. OpenAI, for example, disappointed in August with its GPT-5 model, which failed to meet the company's promises.

The current situation highlights the complexity and inherent challenges of AI development, as well as the need for rigorous risk management and full transparency in the process. The cybersecurity industry and end users eagerly await more details about 'Claude Mythos' and its impact on the digital future.