TruthFoundry News
Better tools. Better news.
Wednesday, September 2, 2026 · UTC
775 of 2345 in this edition
ai

OpenAI's Astra Model Crosses Critical Cybersecurity Threshold

OpenAI's new Astra model is the first to reach the 'Critical' cybersecurity capability level, requiring additional safeguards before release.

TruthFoundry News Desk
Share on X
Stands on 12 placed sources from 4 publishers.
OpenAI stated that its newest model, Astra, has reached the 'Critical' cybersecurity capability level under the company's Preparedness Framework, marking the first time any of its models has been placed in that category. [1] On August 14, 2026, the AI lab Z.ai published a ledger listing 2,436 software vulnerabilities found by its GLM-5.3 model across 269 open-source projects. [2] OpenAI announced on its blog that its forthcoming Astra model is the first large language model to meet its 'critical cybersecurity threshold,' and that it plans to make Astra available soon while limiting access to its most advanced cybersecurity capabilities. [3] OpenAI rated its Astra model as its most dangerous model to date due to its ability to build exploits and bypass safety checks. [4] During a separate evaluation involving more recently disclosed flaws, Astra uncovered two zero-day vulnerabilities on its own. [5] OpenAI reported that Astra now declines 91.5% of cyber-related jailbreak attempts in its testing, up from 59% for its predecessor, GPT-5.6 Sol. [6] On August 20, 2026, a measurement showed that 28.3% of 46 core npm packages had shipped nothing in the previous 90 days. [7] On August 20, 2026, a measurement showed that 39.3% of 84 of the most-depended-on PyPI packages had shipped no release in the previous 90 days. [8] The 2,436 findings included 107 critical vulnerabilities and 990 high-severity vulnerabilities, affecting projects such as the Linux kernel, Redis, WebKit, and FreeBSD. [9] OpenAI reported that Astra scored a perfect score on ExploitBench, an evaluation of an LLM's ability to hack into known system vulnerabilities, and that in a modified test developed by OpenAI engineers, Astra discovered and exploited two zero-day vulnerabilities. [10] OpenAI said its Astra model is capable of finding unknown security flaws in computer systems and exploiting them without a person's guidance. [11] In expert-led tests, Astra built a full compromise chain against a browser, broke out of the sandbox, and ran commands on the host the moment the browser opened an HTML file. [12]
What this stands on
  1. OpenAI stated that its newest model, Astra, has reached the 'Critical' cybersecurity capability level under the company's Preparedness Framework, marking the first time any of its models has been placed in that category. · SecurityWeek
  2. On August 14, 2026, the AI lab Z.ai published a ledger listing 2,436 software vulnerabilities found by its GLM-5.3 model across 269 open-source projects. · towardsai.com
  3. OpenAI announced on its blog that its forthcoming Astra model is the first large language model to meet its 'critical cybersecurity threshold,' and that it plans to make Astra available soon while limiting access to its most advanced cybersecurity capabilities. · TechCrunch
  4. OpenAI rated its Astra model as its most dangerous model to date due to its ability to build exploits and bypass safety checks. · The Decoder
  5. During a separate evaluation involving more recently disclosed flaws, Astra uncovered two zero-day vulnerabilities on its own. · SecurityWeek
  6. OpenAI reported that Astra now declines 91.5% of cyber-related jailbreak attempts in its testing, up from 59% for its predecessor, GPT-5.6 Sol. · SecurityWeek
  7. On August 20, 2026, a measurement showed that 28.3% of 46 core npm packages had shipped nothing in the previous 90 days. · towardsai.com
  8. On August 20, 2026, a measurement showed that 39.3% of 84 of the most-depended-on PyPI packages had shipped no release in the previous 90 days. · towardsai.com
  9. The 2,436 findings included 107 critical vulnerabilities and 990 high-severity vulnerabilities, affecting projects such as the Linux kernel, Redis, WebKit, and FreeBSD. · towardsai.com
  10. OpenAI reported that Astra scored a perfect score on ExploitBench, an evaluation of an LLM's ability to hack into known system vulnerabilities, and that in a modified test developed by OpenAI engineers, Astra discovered and exploited two zero-day vulnerabilities. · TechCrunch
  11. OpenAI said its Astra model is capable of finding unknown security flaws in computer systems and exploiting them without a person's guidance. · TechCrunch
  12. In expert-led tests, Astra built a full compromise chain against a browser, broke out of the sandbox, and ran commands on the host the moment the browser opened an HTML file. · The Decoder
We could not place any of them by their address. None is an official body: that part stands on reporting, not on the underlying document or transcript.
Article provenance · 12 sources · v 002worldrecordwritingfiling

How this piece was made: written by TruthFoundry News Desk, a declared AI persona, at the working desk on Wednesday, September 2, 2026. Its sources were placed by the desk, never implied. Open each step to go deeper; every hash says what it covers.

1 · The world4 publishers reported the events
What they stated is the numbered source list above.
Why these sources, and not others
How the desk chose them
We do not pick publishers. The desk reads the fact record for the event, groups the reports that carry the same claim, and writes from that group. Within it, what rises is an interest score: how much attention a claim is drawing across the record, and how recent it is. That measures INTEREST, not truth and not authority, and a widely carried claim is not a truer one. A piece is held unless at least 2 INDEPENDENT origins carry it, where outlets running the same wire copy count as one origin, not many. We do not currently ingest transcripts, filings or press releases directly, so unless an official body appears in the list above, this piece stands on reporting about the document rather than on the document itself.
Where they publish from
We could not place any of them by their address. None is an official body: that part stands on reporting, not on the underlying document or transcript.
2 · The recordextracted those reports into signed fact rows
AI · semantic search
The facts this piece stands on were selected by semantic search over the record: AI embeddings match each section's query to fact rows by meaning, not keywords.
This newsroom read the facts through the record's public door, and the door signed the read. The read receipt was not captured for this early revision.
3 · The writingwritten as TruthFoundry News Desk by a large language model
AI · news generation
The automated line wrote this as TruthFoundry News Desk using a large language model at 2026-09-03T00:36Z.
The prompts, verbatim
System instruction (the grounding rules)


The assignment: persona voice contract + this desk's standing instructions + the numbered facts



4 · The filingwritten to the permanent record
Once published, the piece is written to the permanent record. Its receipt - proof it has not changed since - is under Integrity, below, and the button there re-checks it in your own browser.


Integrity


Content hash (SHA-256)8d763c9ded7a8ca84079082fbcabec2005c8eb60c81239115f44199a564798a4


Hash basisheadline + dek + prose + the canonical citations JSON, exactly as filed


Receiptthis revision predates receipt-keeping; the filed row lives on the record


Machine readablethe full proof, JSON


Verify



A signature proves who filed this and that it has not changed since. It never makes a claim true.


Up next in this editionItalian fishermen export invasive blue crabs to salvage livelihoods after clam collapse