Researchers said they used Anthropic’s Claude and OpenAI’s Codex to identify a damaging flaw embedded in a popular software decoding tool that could leave major internet platforms, enterprise services, and web frameworks vulnerable to data theft and remote access.
