REPLAY NEWS · in brief
TOUS 4723 Des chercheurs piratent OpenAI avec l aide d une IA d Anthropic
Researchers from the startup Hacktron AI successfully exploited two security vulnerabilities at OpenAI using Anthropic's Claude model. The operation, conducted as part of a cybersecurity research program, allowed them to access employee ChatGPT and Codex accounts and reach the internal development environment. OpenAI, alerted by the researchers, patched the vulnerabilities and awarded them a $6,500 reward.
© REPLAY NEWS — All rights reserved. Substantial reproduction requires permission.