Back to Global AI & Tech NewsData Security & Privacy
Researchers used Anthropic’s Claude to hack into OpenAI
Source: TechCrunch AI
Executive Summary
Independent security researchers successfully used Anthropic's Claude model to break into OpenAI by exploiting vulnerabilities in its community forum software. OpenAI resolved the issues and awarded the startup $6,500 through its bug-bounty program, highlighting growing security pressures on leading AI companies.
In a twist capturing the state of AI security, independent security researchers used Anthropic’s Claude to break into OpenAI, exposing cracks in the ChatGPT-maker's defenses. A three-person security team at startup Hacktron AI carried out the attack as part of an OpenAI bug-bounty program, managing to chain two critical vulnerabilities to gain access to multiple employee accounts.
OpenAI has resolved the issues uncovered by Hacktron, awarding the startup a $6,500 bug bounty. The incident demonstrates how off-the-shelf technology can be used to find vulnerabilities in advanced infrastructure, with experts noting that tools costing $200 a month can enable sophisticated hacks. The entry point was found via a flaw in Discourse, the third-party software powering OpenAI’s community forum.
Confirm and follow the full story at the original source:TechCrunch AI