Cathedral, a months‑old startup founded by ex‑DOGE staff, says it breached Anthropic's internal systems in under 12 hours using AI agents
agents anthropic claude startup
| Source: Techmeme | Original article
Cathedral, a newly founded cybersecurity startup by former DOGE employees, claims it infiltrated Anthropic's internal systems in under 12 hours using AI agents.
Cathedral, a cybersecurity start‑up only months old, says its researchers used AI‑driven agents to infiltrate Anthropic’s internal network in less than 12 hours. The breach, reported by the Washington Post, gave the team access to Slack conversations, meeting transcripts and employee salary data. Cathedral entered the system through Anthropic’s own bug‑bounty program for external security researchers, but the AI‑powered approach reportedly uncovered a path far faster than traditional manual testing.
Anthropic, the creator of the Claude chatbot, has pushed back, stating that the vulnerability was “much smaller” than Cathedral’s claim. The dispute highlights a growing tension between AI developers and security firms that are increasingly leveraging the same generative technologies they aim to protect. If AI agents can map and exploit complex corporate environments in hours, the risk profile for high‑value AI models and the data they process may shift dramatically.
The episode arrives as Cathedral, founded by former staff of the AI‑focused DOGE project, pursues U.S. government contracts to build AI‑enabled offensive and defensive cyber capabilities. The firm raised $160 million at a $1.4 billion valuation in July, backed by Andreessen Horowitz and Sequoia Capital, underscoring the Pentagon’s appetite for private‑sector AI expertise amid rising geopolitical tensions.
Watch for Anthropic’s technical response and any changes to its bug‑bounty scope, as well as potential regulatory scrutiny of start‑ups that blend AI research with military‑grade cyber operations. The incident may also prompt other AI firms to reassess their internal defenses and the role of AI agents in both protecting and probing their systems.
Sources
Back to AIPULSEN