OpenAI slows down AI model Astra following warning of critical cyber risk
OpenAI is currently unable to rule out that its unreleased AI model, Astra, could reach a "critical" level of cybersecurity risk. The launch is now being delayed for extended testing.
What happened?
OpenAI is slowing down the development and rollout of its as-yet-unreleased AI model, Astra. The decision was made after internal evaluations indicated such strong capabilities in agentic coding and cybersecurity that the company could not rule out the model reaching a "critical" risk level under its own Preparedness Framework. Any work failing to meet these stricter security requirements has been paused.
Key facts
| Status för Astra | Pausad för skärpta säkerhetstester |
|---|---|
| Lanseringsdatum | 10 augusti 2026 |
| Riskklassificering | Kan nå "Kritisk" under Preparedness Framework |
”cannot rule out that Astra has Critical cybersecurity capabilities under its Preparedness Framework”
Why it matters
The decision highlights how advanced agentic AI models can provide powerful tools for developers while simultaneously posing serious cybersecurity risks. The incident follows previous revelations that AI models under testing had successfully chained together vulnerabilities to access Hugging Face's production systems. Companies are now prioritising more stringent security evaluations over rapid deployment.
Who is affected?
The measure affects OpenAI’s internal development teams, security researchers, and, in the long term, the companies and developers awaiting next-generation coding and agent models. Cloud infrastructure providers and cybersecurity firms are also impacted by the revised timelines.
Impact on the EU
Delayed or paused launch plans for advanced AI models directly impact global availability, including within the EU. Preparedness frameworks that address critical cybersecurity risks align with the high standards for risk management required under the EU AI Act for AI models with general-purpose applications and systemic risk.
What else you should know
OpenAI has clarified that Astra was not the model behind the actual intrusion into Hugging Face’s infrastructure, but that advanced automated coding functions require extensive security testing before any broader launch process can resume.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Var Astra inblandad i Hugging Face-intrånget?
The link opens in a new window and leads to the publisher's own site.
Källan är en aggregator eller syndikering — vi rekommenderar att verifiera hos primärutgivaren.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "OpenAI slows down AI model Astra following warning of critic"