Anthropic to launch Mythos AI model tomorrow with advanced reasoning features: Report

Anthropic could finally be preparing to release its powerful Mythos AI model to the public. As per a new report by Sources, while the new model will bersion of Mythos, it will be launched under different name.

The report also notes that the public facing version of Mythos will come with significantly more guardrails than the version Anthropic provided to companies with Project Glasswing.

What is Mythos? Why is it relevant?

Mythos is Anthropic's most poweful model to date which it had unveiled a couple of months back but refused to release to the public due to fears of it being exploited for cyber attacks.

The company said that Mythos could surpass all but the most skilled human security researchers in identifying and exploiting software vulnerabilities. So, Anthropic gave the critical infrastructure providers an early access to Mythos to test out their systems under Project Glasswing which initially included Apple, Google, Microsoft among other big names. The project was later expanded to include companies from over 15 countries including India and companies expanded to over 100.

The AI startup has said that its AI model has already discovered thousands of high-severity vulnerabilities across major operating systems, web browsers and other critical software that had sometimes previously gone unoticed for decades.

The powerful model found a 27-year-old flaw in OpenBSD, a 16-year old vulnerability in the widely used FFmpeg video-processing software and another vulnerability in Linux that could allow attackers to gain control of the user's machine.

Meanwhile, Mozilla said that Mythos had help identify and fix 271 vulnerabilities in Firefox, over 10 times the number it found with Claude Opus 4.6.

Anthropic shares how attackers could use Mythos to take advantage of software vulnerabilities:

In a new blogpost, Anthropic said that while the company has been warning about new vulnerabilities found using Mythos in the past few months, the model can also be exploited to take advantage of publicly disclosed vulnerabilities.

The company says these vulnerabilities have been disclosed but not yet patched by the developers and attackers take advatange of this ‘patch gap’

“In some ways, N-days are the more dangerous of the two, because the patch itself provides a roadmap to the bug.” Anthropic wrote

The report by Anthropic suggests that organizations now have much less time to patch their vulnerabilities as attackers could take advantage of powerful AI models to turn known vulnerabilities into working exploits in hours.