Război între AI-uri: Chat GPT, spart cu ajutorul Claude, inteligența artificială a rivalului Anthropic

Cybersecurity researchers accessed OpenAI's ChatGPT accounts using Anthropic's AI, Claude, to identify vulnerabilities. This exercise revealed significant security risks associated with the Discourse platform. The Hacktron AI team demonstrated that the vulnerability could potentially expose vast amounts of data from various services. They initially struggled with Claude Opus 4.8 but succeeded with the newly released Opus 5, which quickly identified an effective exploitation method. OpenAI confirmed the incident and has since patched the vulnerabilities, thanking the researchers for their findings.