OpenAI 'ethically hacked' with help of Anthropic's Claude chatbot (www.theguardian.com)

Cybersecurity researchers from Hacktron AI successfully "ethically hacked" OpenAI with assistance from Anthropic’s Claude chatbot, exposing vulnerabilities within the company’s systems. By leveraging Claude’s coding capabilities, the team accessed ChatGPT accounts of OpenAI employees and initiated a harmless “pull request” to alter code on OpenAI’s GitHub repository. While they identified significant potential access to OpenAI’s data, the researchers reported the findings to the company through its bug bounty program without downloading sensitive information. OpenAI acknowledged their discoveries and addressed the exploited vulnerabilities. This incident underscores the increasing sophistication of AI-driven hacking techniques, where advanced tools like Claude and OpenAI's own GPT-5.6 model significantly expedite what once required extensive resources and time to execute. The successful hack highlights ongoing concerns in the AI/ML community regarding cybersecurity risks posed by fast-evolving technology. The situation may also amplify calls for responsible AI development, as both Anthropic and OpenAI endorse a cautious approach amidst escalating fears of unregulated AI advancements.