Anthropic says its latest AI model can uncover vulnerabilities in software security

Anthropic said Tuesday that its yet-to-be-released artificial intelligence model called Claude Mythos has proven extremely adept at uncovering software weaknesses.

Mythos has uncovered thousands of vulnerabilities in commonly used applications that have no patch or fix, prompting the San Francisco-based AI startup to form an alliance with cybersecurity specialists to strengthen defenses against hacking attacks and prevent widespread spread.

“We have a new model that we are specifically not releasing to the public,” Mike Krieger of Anthropic Labs said at a HumanX AI conference in San Francisco.

Instead, Anthropic has cybersecurity specialists and engineers in the open source community working with Mythos to use the model as a defensive weapon, “to prematurely weaponize them, so to speak,” Krieger explained.

Leaps in the capabilities of AI models come with concerns that hackers could use such tools to figure out passwords or crack encryptions intended to ensure data security.

The oldest of…