OpenAI's Astra model hits Critical cybersecurity threshold
Astra is the first OpenAI model to reach the Critical cybersecurity capability threshold under the Preparedness Framework, scoring 100% on ExploitBench and finding 2 zero-days in evals. OpenAI is adding safeguards before release, which is "coming soon."
How this story unfolded
4 weeks · 24 reports · 40 community posts · 64 of 70 shown
- Aug 3
- Aug 6
- Aug 7
Responding to the next frontier of critical cyber capabilitiesopenai.com
OpenAI Pauses Some Work on New Astra Model Over Cyber Concernsbloomberg.com
OpenAI puts the brakes on a new model because it’s supposedly too powerfultheverge.com
The AI model OpenAI won’t release yet — and what it found in testingthenewstack.io
- Aug 10
OpenAI's Next AI Model Astra Shows Cyber Performance Strong Enough to Trigger Pausethehackernews.com
OpenAI tightens controls on its new model over cybersecurity risks, as AI security debate intensifiescnbc.com
OpenAI’s Upcoming Astra Model Raises Autonomous Cyberattack Concernssecurityweek.com
OpenAI locks down Astra after model raises first-ever critical cyber capability fears
- Aug 13
- Aug 17
- Aug 18
- Aug 20
- Aug 29
- Aug 30
- Aug 31
- Sep 1
OpenAI Is About to Release Its First AI Model With ‘Critical’ Cyber Abilitieswired.com
Open AI’s Astra model is on the way—and very good at breaking into computer systemstechcrunch.com
- Sep 2
OpenAI Astra and Looped Transformerssebastianraschka.com
OpenAI's Astra Becomes Its First AI Model With 'Critical' Hacking Abilitiesdecrypt.co
OpenAI by email
Get an email when OpenAI has news
No news that day, no email.
More stories today
- Meta builds AI 'second brain' that learns from experts
- Snowflake shares surge 22% on AI coding momentum
- Gary Marcus critiques Musk's AI prediction shift
- Lutnick says Anthropic has patched relations with US government
- HPE lifts sales forecast on AI server demand