← Back to feed
ai$MSFT$AMZNproduct launch1 source · 6d ago

GPT-6 Astra Scores 100% on ExploitBench as OpenAI Blocks PoC Exploit Requests

OpenAI has released GPT-6 Astra, an AI model that achieved a 100% score on the ExploitBench benchmark while implementing safety restrictions to prevent the generation of proof-of-concept exploits.

Affected◉ United States▪ Amazon Web Services▪ ExploitBench▪ GPT-6 Astra▪ Microsoft Azure▪ MS-ISAC▪ OpenAI
⌗ 1 outlet · 1 origin · Balanced
India × 1

No Reader read of this story yet.

Perspectives

The story's competing narratives, side by side — grouped by stance, with every outlet's origin and affiliation visible.

Perspective analysis pending — it generates as coverage from more origins arrives.

What to expect

First-order impacts with their likely second-order effects — direction and horizon per node.

Impact analysis pending.

Sources (1)

GPT-6 Astra Scores 100% on ExploitBench as OpenAI Blocks PoC Exploit Requests — Prism