jasdooit

Anthropicis developing a new advanced artificial intelligence (AI) system that “presents extraordinary cybersecurity dangers,” as reported by a disclosure from the organization.

Last week, a data breach disclosed that Anthropic was developing a new, advanced model, which the AI company has now confirmed is named “Claude Mythos.”

An Anthropic spokesperson told Fortuneit signifies a “significant shift” in artificial intelligence and represents the “most powerful” model it has developed to date, featuring “substantial improvements in logical thinking, programming, and network security.”

The technology might be powerful enough to become a hacker’s ideal tool. Cybersecurity shares declined after the rumors about Anthropic surfaced.

Meanwhile, Anthropic is privately informing senior government officials that Mythos significantly increases the likelihood of large-scale cyberattacks in 2026,Axios reported.

Here is all the information we have regarding Mythos.

How did the disclosure come to light?

A setup mistake, or a mistake made by a person, in Anthropic’s content management system unintentionally revealed a draft blog post that outlined the model.

The blog, referenced by Fortune, which was “accessible in an unsecured and easily searchable data repository,” claimed that the new model “introduces extraordinary cybersecurity dangers.”

What can Mythos do?

Mythos is believed to be a component of Capybara, which represents a new level of AI models.

Fortune reported that Antropic mentioned in the document, “‘Capybara’ is a new designation for a higher level of model: bigger and smarter than our Opus models—which were, up until now, our strongest.”

The model’s enhanced abilities seem to indicate that AI systems are capable of addressing more complex issues with greater independence and accuracy.

Capybara and Mythos seem to point to the same foundational model.

What impact will it have on cyber security?

As AI firms introduce enhanced models, AI agents—capable of learning to act and think independently without human intervention—also improve.

Cybercriminals can thus conduct several hacking operations simultaneously, making it harder to defend against.

At the same time, employees are utilizing AI assistants that might unintentionally link to their own work systems, providing cybercriminals with an entry point.

Furthermore, personal identities can now be more easily compromised due to AI, according to Palo Alto Networks.executiveinformed Euronews Next in March.

When can we expect to see Mythos?

Mythos requires a significant amount of computational power and is costly to operate. Anthropic mentioned that they are focusing on improving its efficiency prior to a broad launch.

Euronews Next has contacted Anthropic for a response.