Anthropic's Claude helps hack OpenAI via $200/month tools
researchers used anthropic's claude to break into openai, exposing vulnerabilities, for just $200 a month. this incident highlights growing ai security risks.
independent security researchers at hacktron ai used anthropic's claude model, specifically opus 5, to hack into openai's systems. they exploited vulnerabilities in discourse and libheif to gain access to openai employee chatgpt and codex accounts. the attack was part of an openai bug-bounty program and resulted in a $6,500 award. this incident highlights the growing capabilities of ai models in cybersecurity and the potential risks to even advanced companies.