AI companies spend enormous amounts of time convincing us their technology is safe and secure. But what happens when those same companies become targets of increasingly sophisticated AI-powered attacks? Anthropic, the company behind Claude, is now facing that question. According to a Washington Post report, cybersecurity startup Cathedral discovered vulnerabilities that let its researchers access sensitive information in Anthropic’s internal systems.
The researchers used AI agents to identify potential entry points in less than 12 hours. Interestingly, they didn’t need to crack Claude itself. Instead, they exploited weaknesses in third-party business applications that Anthropic uses for everyday operations. Anthropic’s security testing program disclosed the findings, which allows outside researchers to identify and report vulnerabilities.
The researchers found more than they expected
Once inside, Cathedral’s researchers uncovered routes to information that companies would generally want to keep private. This included employee salary details, job offers, meeting transcripts, and internal Slack conversations about Anthropic’s advanced AI models. The most concerning part is how the researchers got there. Rather than attacking Anthropic’s core AI infrastructure, they targeted ordinary business software used for administrative tasks and customer management.
It’s a reminder that even companies developing some of the world’s most advanced technology can be exposed through the everyday tools their employees rely on. Anthropic acknowledged that the researchers accessed a limited number of internal sales, human resources, and IT systems. However, it emphasized that the incident did not compromise customer conversations with Claude, the company’s underlying AI models, or their source code and model weights. The company also said its security systems detected some suspicious activity before Cathedral submitted its findings. It then blocked access and disabled the affected applications until their vendors addressed the vulnerabilities. Anthropic maintains it has found no evidence that anyone else exploited the same weaknesses.
Why this discovery is bigger than Anthropic
The incident comes at a particularly sensitive time for the AI industry, with governments increasingly concerned about protecting advanced AI research from foreign intelligence agencies and cybercriminals. Cathedral, founded by former members of the US government’s DOGE initiative, has secured federal cybersecurity contracts and attracted a $1.4 billion valuation. Its findings have also caught the attention of US officials, although Anthropic disputes the startup’s characterization of the potential national security risks.
Another important distinction is that this was authorized security research, not evidence of a successful attack by malicious hackers. Anthropic also says Cathedral did not fully comply with its disclosure program’s requirements. Still, the broader concern is difficult to ignore. As AI becomes better at identifying and exploiting software vulnerabilities, even seemingly harmless business applications could become attractive targets. For companies building powerful AI systems, securing the technology itself might be only half the battle.
