OpenAI Previews Astra Reaching Critical Cybersecurity Threshold
OpenAI reports its first model to hit the Critical cybersecurity threshold under the Preparedness Framework.
OpenAI announced that Astra is the first of its models to reach the Critical cybersecurity capability threshold under the Preparedness Framework. The company is preparing the model for release and previewed its evaluation process along with stronger safeguards. It plans to start with more limited access to the most advanced cybersecurity features before expanding availability. Posts from OpenAI engineers note performance gains over earlier versions on ExploitBench and discuss rollout steps.
Combined views
1.7M
10 posts, first seen 10h ago
Reactions
10 postsOP 
OpenAI@OpenAIAs we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly accessible. Astra represents a significant advance in cybersecurity capability, reaching the Critical threshold under our Preparedness Framework. We're previewing how we evaluated…
- likes: 10.2K
- replies: 703
- bookmarks: 1.4K
- reposts: 835
FO fouad@fouadmatinAstra is both significantly more capable and more token efficient compared to 5.6 Sol, as seen in port of ExploitBench. We plan to make Astra available soon, with its most advanced cybersecurity capabilities initially more limited, then iteratively expand access more broadly.
- likes: 936
- replies: 42
- bookmarks: 106
- reposts: 57
CH Chubby♨️@kimmonismusAll the problems seem to have been resolved. GPT-Astra is coming "soon." Rumor has it, this Thursday. The fact that Fable 5.1 was released today makes this even more likely. And frankly, I'm much more excited for GPT-Astra than I am for Fable 5.1. So far, I honestly haven't…
- likes: 919
- replies: 49
- bookmarks: 55
- reposts: 38
BD Brendan Dolan-Gavitt@moyixI’d love to read the solutions in that 1.3% - given that (speculating slightly) the honeypot tasks are the ExploitGym problems no model has ever solved there must be some really cool tricks in there
- likes: 16
- replies: 0
- bookmarks: 4
- reposts: 1
XQ Xiangyu Qi@xiangyuqi_ptonAstra achieves a full 100% success rate on ExploitBench, so we had to build an internal refresh using newly disclosed vulnerabilities from June through August that fall after the model’s knowledge cutoff. On this refreshed benchmark, Astra remains dramatically stronger than…
- likes: 1.4K
- replies: 40
- bookmarks: 248
- reposts: 95
FO fouad@fouadmatinAstra is both significantly more capable and more token efficient compared to 5.6 Sol, as seen in port of ExploitBench. We plan to make Astra available soon, with its most advanced cybersecurity capabilities initially more limited, then iteratively expand access more broadly.
- likes: 936
- replies: 42
- bookmarks: 106
- reposts: 57
