← Founder Notes
Archive

Gpt-6 astra hit 100% on exploitbench and openai bumped it to critical-cyber under their…

Yethikrishna ROriginal on Threads

gpt-6 astra hit 100% on exploitbench and openai bumped it to critical-cyber under their preparedness framework. the dev timeline keeps calling it agi.

the exploit dev kit will ship first.

Context

OpenAI's Path to Astra post of 1 September 2026 says it now believes Astra meets the Critical cybersecurity capability threshold under its Preparedness Framework. A search excerpt of the same page says Astra achieved a perfect score of 100 percent on ExploitBench, a benchmark of developing exploits from known vulnerabilities. I did not read that line in the page text I fetched.

One outlet says the 100 percent rests on 41 already-known bugs, which I have not checked against OpenAI's page. The New Stack quotes Brockman saying it is not unreasonable to feel that we are now in the AGI era.

How it compares

Bumped to critical-cyber matches OpenAI's wording. The prediction that the exploit dev kit ships first is the author's, and OpenAI's page says advanced cyber access goes to testers and defenders first.

Related work

Watch next

  • The system card's cybersecurity section for the ExploitBench scope.

Sources

  1. The path to Astra (OpenAI)openai.com
  2. GPT-6 Astra system card (OpenAI)deploymentsafety.openai.com
  3. OpenAI launches GPT-6 Astra benchmarks (The New Stack)thenewstack.io

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 4 September 2026 at 16:03 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/gpt-6-astra-hit-100-on-exploitbench-and-Dc3LppTjYCs" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="Gpt-6 astra hit 100% on exploitbench and openai bumped it to critical-cyber under their…"></iframe>

More notes