A fresh cybersecurity alarm is sounding in the artificial intelligence industry as Anthropic investigates a possible breach involving its newly launched Mythos model, raising urgent questions about third-party security risks and the future of AI-powered vulnerability detection.

The model, still in its early rollout phase, was designed to help organizations identify software weaknesses before hackers can exploit them. Now, that same system is at the center of a security probe after reports of unauthorized access surfaced.

Investigation Underway After Vendor Access Concerns

Anthropic, the company behind the Claude chatbot, is examining reports of potential unauthorized access to Mythos through a third-party vendor environment, according to a company spokesperson speaking to CBS News via email.

The company confirmed it works with a limited number of external vendors to support AI model development and stressed that, so far, no evidence shows breaches beyond those vendor systems.

It also said there is no indication that Anthropic’s internal systems have been compromised.

The investigation was publicly confirmed Wednesday, one day after Bloomberg reported that a small group of unauthorized users may have accessed the tool, citing a person familiar with the situation.

Limited Release, High Stakes Security Tool

Anthropic introduced Mythos in April under a controlled program called Project Glasswing. The rollout was intentionally restricted to a small number of major technology and financial organizations, including Amazon, Apple, Cisco, JPMorgan Chase and Nvidia.

The company positioned Mythos as a high-performance tool for identifying software vulnerabilities, designed to outperform competing systems in cybersecurity detection.

However, the restricted access was also driven by concern. Anthropic limited distribution due to fears the model itself could be targeted or exploited by hackers seeking to probe sensitive infrastructure.

The goal of Project Glasswing is to strengthen defenses before malicious actors can gain access to similar AI tools or weaponize them against critical systems.

Growing Global Concerns Over AI Cybersecurity Risks

The potential consequences of a breach go far beyond corporate systems. Federal officials, cybersecurity experts, and global institutions such as the International Monetary Fund have all warned about the risks if advanced AI security tools fall into the wrong hands.

Experts caution that systems like Mythos could be used not only for defense but also to probe and exploit weaknesses in banking networks, hospitals, government infrastructure, and other essential services.

The dual-use nature of the technology has intensified concerns that AI could dramatically escalate the speed and scale of cyberattacks.

Expert Warning on AI-Driven Threats

"We need to prepare ourselves, because we couldn't keep up with the bad guys when it was humans hacking into our networks," Alissa Valentina Knight, CEO of cybersecurity AI company Assail, previously told CBS News." We certainly can't keep up now if they're using AI because it's so much devastatingly faster and more capable.

As the investigation continues, the incident underscores a growing reality in cybersecurity: the same tools built to defend digital systems may also become some of the most powerful assets in future cyber warfare.