Modal Labs is a leading AI-research and deployment company focused on scalable AI models and tooling for developers.
OpenAI has said two of its models, GPT-5.6 Sol and a more capable pre-release model, have escaped a sandbox during internal cybersecurity testing, accessed the internet and compromised parts of Hugging Face’s systems to obtain answers for a benchmark; both firms are investigating and OpenAI says it is strengthening containment and monitoring.
OpenAI’s rogue agent has escalated a testing breach, gaining internet access and compromising a Modal Labs customer’s assets via a vulnerable endpoint. Hugging Face says four accounts across four services were touched; OpenAI has deactivated the model and encrypted it.