# OpenAI Halts Work on Next Model as ‘Critical’ Cyber Risk Tests AI Safety Promises

*Monday, August 10, 2026 at 6:12 AM UTC — Hamer Intelligence Services Desk*

**Published**: 2026-08-10T06:12:59.074Z (4h ago)
**Category**: cyber | **Region**: Global
**Importance**: 8/10
**Sources**: OSINT
**Permalink**: https://hamerintel.com/data/articles/13806.md
**Source**: https://hamerintel.com/summaries

---

**Deck**: OpenAI has paused some work on its upcoming AI model, Astra, after internal tests suggested it could enable “critical” cyber operations, triggering tighter security controls. The move turns an abstract debate over AI risk into an immediate question for governments, hackers and regulators: what happens when a model is deemed too capable to release on schedule?

OpenAI has put parts of its next-generation AI project on hold after internal tests showed the system could plausibly enable “critical” cyber capabilities, a rare public acknowledgment that a leading developer is slowing its own progress over security fears. The pause, reported on 10 August, affects some work involving Astra, the code name for OpenAI’s forthcoming model. According to the company’s account, internal evaluations suggested Astra might be able to assist in cyber operations at a level serious enough that OpenAI said it could not rule out “Critical” capabilities, a label that triggered stricter internal safeguards. In response, the firm has tightened access and controls while it revisits how to…

---

*Full article available with Hamer Intel Pro — https://hamerintel.com/pricing*