Researchers Test OpenAI Defenses With Help From Anthropic's Claude

Cybersecurity researchers say they gained access to OpenAI systems during an authorized test, using Anthropic's Claude chatbot as part of the effort.

The team at US startup Hacktron AI compromised ChatGPT accounts belonging to OpenAI employees, a process that opened the door to the company's software cache and potentially more. "The scope of what we could theoretically access was huge," the researchers said.

Claude, which can generate code for hackers, was used first to reach ChatGPT accounts through an OpenAI staff forum hosted on Discourse. The team then submitted a harmless pull request to OpenAI's service on GitHub.

Hacktron reported the findings to OpenAI under a program that rewards ethical hackers for testing its systems. The researchers said they had access to code in the GitHub repository but did not download it. They added that they largely relied on OpenAI's own models to carry out the operation, which was first reported by the Wall Street Journal.

An OpenAI spokesperson said: "We thank the researchers for contacting us and sharing their findings," adding that the exploited vulnerabilities had been addressed.

Hacktron said AI tools made a once-complex hacking task far easier and sharply cut the time needed to plan and execute an attack. "Work that once required a well-resourced team and months of effort can now be compressed into days," the company said. It received a $6,500 payment from OpenAI through its bug bounty program.

The incident is the latest safety episode at OpenAI, which disclosed in July that a "swarm" of autonomous agents powered by its technology had hacked the AI startup Hugging Face during a cybersecurity test. This week the San Francisco company revealed six more examples of "unexpected or concerning" actions by its technology and warned that development could not continue at "maximum speed for much longer."

Anthropic made a fresh call at the weekend for a slowdown in AI development, backed by OpenAI, Google DeepMind and Elon Musk. Anthropic repeated warnings that unrestrained AI development poses an existential threat, concerns some experts dispute. Donald Trump has rejected calls for a slowdown, citing the need to stay ahead of China's AI industry and dismissing "negative forces ... bringing up things that won't happen."

Source: The Guardian