AI Round-up: 29 September 2026

OpenAI's reported loss-of-control incidents and criticism of frontier labs' safety claims dominate the news, while NVIDIA says its containment tool can stop rogue agents in milliseconds. The debate centres on whether safeguards can keep pace with increasingly autonomous systems.

Practical demonstrations show ChatGPT modelling in Autodesk Fusion, Microsoft's Copilot Agent Builder and Managed Runtime, and an LLM wiki built with Elasticsearch. Cursor coverage examines automated rollouts and security review.