White Hat Hackers Breached OpenAI Using Anthropic's Claude in Under 72 Hours, TechRadar Reports

The account, published at 10:49 UTC by TechRadar Pro, concerns an intrusion carried out by security researchers rather than criminal attackers. The headline names Claude, Anthropic's model, as the tool used against OpenAI, the developer of ChatGPT and Anthropic's principal commercial rival in the frontier model market. TechRadar did not, in the material available, specify which OpenAI systems were reached, what data or access was obtained, or how the researchers moved from the model to the target.

The report is credited to Benedict, a senior security writer at TechRadar Pro whose beat covers geopolitics, cyber-warfare and business security. The biography attached to the article describes him as specializing in state-sponsored threat actors, advanced persistent threat groups and the protection of critical national infrastructure, and as holding a master's degree in security, intelligence and diplomacy from the University of Buckingham Centre for Security and Intelligence Studies, along with a bachelor's degree in politics with journalism.

No statement from OpenAI or Anthropic appeared in the available material, and neither company had publicly confirmed or disputed the account. TechRadar also did not indicate whether the researchers operated under a formal agreement with OpenAI, or whether the findings were reported to the company before publication.

The phrase "in less than 72 hours" is the report's central claim, and TechRadar frames that speed as evidence of the rate at which AI systems are gaining capability, rather than as a one-off incident. Because the article as provided carries the writer's background rather than a full technical write-up, key questions — the scope of the intrusion, the vulnerability or technique used, OpenAI's response, and any disclosure timeline — remain unanswered in that source.