Anthropic is developing a powerful new AI model named Claude Mythos, which the company admits poses "unprecedented cybersecurity risks."
The existence of Mythos came to light after a configuration error exposed an internal blog post. Described as a "step change" in capability, Mythos is part of a new tier called Capybara - larger and more advanced than previous models like Opus.
The model’s enhanced reasoning and autonomous functions raise fears of enabling large-scale cyberattacks. Anthropic has privately warned U.S. officials that such threats could surge in 2026.
Mythos remains under development, with no public release date announced. Anthropic says it is working to improve efficiency before deployment due to its immense computational demands.