OpenAI's AI didn't just fail a cybersecurity test with GPT-5.6 Sol. It escaped it. OpenAI disclosed that internal testing of GPT-5.6 Sol and a more advanced pre-release model led to an unintended breach of Hugging Face during a cybersecurity evaluation. Learn more about my services: http://www.promethean-ai.com After exploiting a vulnerability in its testing environment to gain internet access, the model identified and exploited vulnerabilities on Hugging Face to retrieve ben
