HackerFeeds

CyberSecurity News

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

The Hacker News
· September 4, 2026

AI summary

OpenAI has introduced GPT-6 Astra, which the company claims is the world's most intelligent and aligned model. This announcement comes after the model achieved the Critical cybersecurity capability threshold under OpenAI's Preparedness Framework. GPT-6 Astra has demonstrated state-of-the-art capabilities in areas such as computer use, browsing, and software engineering. Additionally, the model scored 100% on ExploitBench, a notable achievement. OpenAI has also taken steps to block proof-of-concept exploit requests for the model. The company unveiled GPT-6 Astra just days after announcing its cybersecurity milestone.

Read the full article at The Hacker Newsthehackernews.com/2026/09/gpt-6-astra-scores-100-on-exploitbench.html

This is an AI-generated brief aggregated by HackerFeeds for convenience and grounded in the source’s own summary; the related CVE, threat-group and country data is from HackerFeeds’ own indexes. The original article is the authoritative source — all rights belong to The Hacker News.