AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense
Back to Home
tech

AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense

October 28, 2026128 views2 min read

Cybersecurity startup Cathedral used AI agents to hack into Anthropic's internal systems, exposing sensitive data and highlighting vulnerabilities in even the most advanced AI labs.

In a stark reminder of the evolving cybersecurity landscape, a cybersecurity startup named Cathedral has successfully breached the internal systems of Anthropic, a leading AI company known for developing some of the most advanced AI models for cyberdefense. The breach was carried out using AI agents, demonstrating how even the most sophisticated AI labs are not immune to attacks through their third-party software dependencies.

Method and Scope of the Attack

The attack was executed as part of a formal bug bounty program, where Cathedral utilized AI agents to infiltrate Anthropic’s systems. These agents exploited vulnerabilities in third-party software, gaining access to sensitive data including sales, human resources, and IT-related information. While Anthropic confirmed that the breach allowed access to these data sets, the company has not disclosed the full extent of the compromise, raising questions about whether deeper system infiltration occurred.

Implications for AI Security

This incident highlights a critical vulnerability in the AI industry: the reliance on external tools and platforms. Even organizations with strong internal security protocols can be compromised through weak links in their software supply chains. As AI systems become more integrated into critical infrastructure, such breaches could have far-reaching consequences, especially when targeting companies that specialize in cybersecurity and AI defense.

Security experts are now urging AI labs and tech companies to reassess their third-party software vetting processes and to invest more heavily in securing their entire digital ecosystem, not just internal systems. The use of AI agents in this attack also underscores the growing sophistication of cyber threats and the need for adaptive, AI-driven defense mechanisms.

Conclusion

The breach of Anthropic by Cathedral is a wake-up call for the AI industry. As AI becomes increasingly central to cybersecurity, the need for robust, multi-layered defenses is more urgent than ever. Organizations must move beyond traditional security models and embrace a more dynamic, proactive approach to protect against both human and AI-driven threats.

Source: The Decoder

Related Articles