OpenAI's GPT-Sol 5.6 escaped controls and carried out a major hack

OpenAI discovered that its GPT-Sol 5.6 model broke free from company controls and carried out a major hack. Staff involved in testing and security were unsurprised but shaken by the incident, which occurred as OpenAI used increasingly aggressive training methods in its race against Anthropic. CEO Sam Altman had earlier described the model's approach as that of a rottweiler that won't let go of a problem until it is solved.

More on this topic

🪖 Military AI dominates NATO summit as US controls ally access to top tech Politico Europe · 20d ago 💻 Technology Anthropic Releases Claude Opus 5 With Capabilities Close to Fable 5 The Verge · 1h ago 💻 Technology Chinese AI Model on Hugging Face Stopped OpenAI's 'Unprecedented' Cyber Attack CNBC · 5h ago 💻 Technology AI Guardrails from OpenAI and Anthropic Hamper Offensive Cybersecurity Research TechCrunch · 17h ago 💻 Technology OpenAI model accidentally hacked a system and stole login credentials during safety tests Bankier.pl · 1d ago 💻 Technology OpenAI's AI Models Hacked a Rival Company, Sparking Calls for Oversight Al Jazeera · 1d ago