# AI Capability Growth Widens the Safety and Governance Gap

> OpenAI says its GPT-6.2 Astra model hit a key self-improvement milestone, raising concerns that safety rules are not keeping up.

- Type: AI news
- Page: https://completeaitraining.com/news/ai-capability-growth-widens-the-safety-and-governance-gap/
- Published: 2026-11-21
- Updated: 2026-11-20
- Author: Jeroen Erne
- Jobs: Education, Executives and strategy, Government, Human resources, Management, Science and research, Writers
- Reading time: 1 min

OpenAI has reported that its model GPT-6.2 Astra reached a high threshold for AI self-improvement. The announcement comes as researchers and industry specialists continue to debate the expected pace of progress toward superintelligence.

The development has drawn fresh attention to the gap between the speed of technical capability growth and the maturity of safety and governance structures. Experts are urging boards to establish clear internal controls before further thresholds are crossed. These controls include defining specific capability triggers that would prompt a review, designating who holds release authority, and requiring independent testing of systems prior to deployment.

---

This is the Markdown form of https://completeaitraining.com/news/ai-capability-growth-widens-the-safety-and-governance-gap/. The whole site is mapped at https://completeaitraining.com/llms.txt.
