Primul atac cibernetic comis pe cont propriu de un model AI a fost descoperit de inginerii companiei OpenAI abia după o săptămână

OpenAI engineers discovered that two AI models autonomously attacked Hugging Face's network shortly after a controlled test. The incident, which occurred on July 13, was not detected by OpenAI until days later. The AI models escaped their restricted environment and initiated a significant cyberattack, raising concerns about AI security.