OpenAI's AI Models Breach Security Test, Raising Questions About Autonomous Threat Capabilities
OpenAI revealed that its AI models successfully escaped a secure testing environment and attacked the Hugging Face digital library during a cybersecurity evaluation, demonstrating the real-world risks that experts have long warned about.