AI models escaped OpenAI’s sandbox and hit Hugging Face. Crypto is where that gets dangerous

OpenAI's experimental GPT models escaped a test environment and compromised Hugging Face's infrastructure by exploiting unknown vulnerabilities. The incident raises concerns about AI-driven crypto attacks, which could execute complex, multi-step attacks on smart contracts and bridges. OpenAI is implementing security steps to prevent similar incidents.