{ "story_id": "ade6d672d4c735ec91e3de3c2c1df7ec", "desk": "drm3", "revision": 1, "published_at": "2026-09-03T20:50:00.000Z", "content_hash": "9f597f87d0ed40354c42f16bd1d5a8a777c559cc2e1713f2dbc55d07aa0f5a80", "hash_basis": "sha256 over `headline\\ndek\\nprose`, plus `\\n` + the canonical citations JSON when any source is placed, plus `\\n#blog` for blogs", "basis": { "headline": "OpenAI's Astra Model Crosses Critical Cybersecurity Threshold", "dek": "OpenAI's new Astra model is the first to reach the 'Critical' cybersecurity capability level, requiring additional safeguards before release.", "prose": "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]\n\nGreg Brockman stated that Astra meets the bar for artificial general intelligence, meaning it can match or exceed human capabilities across various tasks. [^2]\n\nOpenAI released GPT-6 Astra on Thursday, which president Greg Brockman called a 'generational leap in capability' and the arrival of artificial general intelligence. [^3]\n\nOpenAI rated its Astra model as its most dangerous model to date due to its ability to build exploits and bypass safety checks. [^4]\n\nOpenAI launched GPT-6 Astra, a new artificial intelligence model described by the company as the world's most intelligent and aligned. [^5]\n\nIn testing described by the company, Astra achieved a perfect score on ExploitBench, a benchmark that measures a model's ability to turn known vulnerabilities into working exploits. [^6]\n\nOpenAI 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. [^7]\n\nOpenAI stated that the 'Critical' classification requires additional safeguards before the model can be released. [^8]\n\nAstra built a browser exploit that escaped the sandbox and ran commands on the host during testing with security experts. [^9]\n\nThe Critical cybersecurity threshold is reserved for models capable of finding vulnerabilities and developing exploits with far less human assistance. [^10]\n\nGPT-6 Astra is the first model OpenAI has designated 'critical' under its Preparedness Framework for cybersecurity. [^11]\n\nOpenAI acknowledged that Astra is the first system the company has rated capable of autonomously hacking well-protected systems without human guidance. [^12]", "cited": "[{\"statement\":\"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.\",\"source\":\"SecurityWeek\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T10:38:13.000Z\",\"publisher_count\":1,\"sources\":[\"SecurityWeek\"]},{\"statement\":\"Greg Brockman stated that Astra meets the bar for artificial general intelligence, meaning it can match or exceed human capabilities across various tasks.\",\"source\":\"Decrypt\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T19:24:24.000Z\",\"publisher_count\":1,\"sources\":[\"Decrypt\"]},{\"statement\":\"OpenAI released GPT-6 Astra on Thursday, which president Greg Brockman called a 'generational leap in capability' and the arrival of artificial general intelligence.\",\"source\":\"Decrypt\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T19:24:24.000Z\",\"publisher_count\":1,\"sources\":[\"Decrypt\"]},{\"statement\":\"OpenAI rated its Astra model as its most dangerous model to date due to its ability to build exploits and bypass safety checks.\",\"source\":\"The Decoder\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T14:20:28.000Z\",\"publisher_count\":1,\"sources\":[\"The Decoder\"]},{\"statement\":\"OpenAI launched GPT-6 Astra, a new artificial intelligence model described by the company as the world's most intelligent and aligned.\",\"source\":\"9to5Google\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T20:50:00.000Z\",\"publisher_count\":1,\"sources\":[\"9to5Google\"]},{\"statement\":\"In testing described by the company, Astra achieved a perfect score on ExploitBench, a benchmark that measures a model's ability to turn known vulnerabilities into working exploits.\",\"source\":\"SecurityWeek\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T10:38:13.000Z\",\"publisher_count\":1,\"sources\":[\"SecurityWeek\"]},{\"statement\":\"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.\",\"source\":\"SecurityWeek\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T10:38:13.000Z\",\"publisher_count\":1,\"sources\":[\"SecurityWeek\"]},{\"statement\":\"OpenAI stated that the 'Critical' classification requires additional safeguards before the model can be released.\",\"source\":\"SecurityWeek\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T10:38:13.000Z\",\"publisher_count\":1,\"sources\":[\"SecurityWeek\"]},{\"statement\":\"Astra built a browser exploit that escaped the sandbox and ran commands on the host during testing with security experts.\",\"source\":\"thenewstack.io\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T20:13:54.000Z\",\"publisher_count\":1,\"sources\":[\"thenewstack.io\"]},{\"statement\":\"The Critical cybersecurity threshold is reserved for models capable of finding vulnerabilities and developing exploits with far less human assistance.\",\"source\":\"thenewstack.io\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-02T20:13:54.000Z\",\"publisher_count\":1,\"sources\":[\"thenewstack.io\"]},{\"statement\":\"GPT-6 Astra is the first model OpenAI has designated 'critical' under its Preparedness Framework for cybersecurity.\",\"source\":\"Decrypt\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T19:24:24.000Z\",\"publisher_count\":1,\"sources\":[\"Decrypt\"]},{\"statement\":\"OpenAI acknowledged that Astra is the first system the company has rated capable of autonomously hacking well-protected systems without human guidance.\",\"source\":\"Decrypt\",\"instrument\":\"News\",\"claim_key\":null,\"published_at\":\"2026-09-03T19:24:24.000Z\",\"publisher_count\":1,\"sources\":[\"Decrypt\"]}]", "kind": "news" }, "receipt_verify": "Ed25519 over the dot-joined string `slice_hash.cursor_from.cursor_to.view.view_version.row_count`; public_key and sig are base64url of the raw 32-byte key / 64-byte signature", "receipt": null, "receipt_note": "this revision predates receipt-keeping (before v0.37.0); the filed row lives in the record", "generation_chain": { "wire": { "stream": "fountain_news", "story_id": "aa57285abcf5242326fa60fc64da158c", "thread_id": "4a1b71b93ea108114c2076f346ca8500", "thread_label": "ExploitBench", "novelty": "UPDATE", "content_hash": "61eb8afb0341b220c80972972bd3cebd1079532edb0a74f04cac04f334f24d70", "last_published_at": "2026-09-03T20:50:00.000Z", "read_receipt": { "slice_hash": "3528f573095b38c4abfe7176deb5e225e82c1af3c7dc91680410a8795d94b55d", "cursor_from": "eyJ0cyI6IjIwMjYtMDktMDNUMjA6MTc6MDAuMDAwMDAwWiIsImlkIjoiNWI1MmIxMjk3M2NhODBmY2NiMjk5ZDgyMWQ5MjljZmEiLCJ2IjoiMSJ9", "cursor_to": "eyJ0cyI6IjIwMjYtMDktMDNUMjA6NTA6MzEuMDAwMDAwWiIsImlkIjoiNTFkOTVlNTZjYmQ0MGI5YmU1MWVjMjY3ZjU4ZWI2YWEiLCJ2IjoiMSJ9", "view": "v_fountain_news", "view_version": "1", "row_count": 100, "window_days": 3, "bytes_scanned": 12166856, "credits": 8, "price_per_100_rows": 8, "sig": "4tF4JvaY1J1xDk_C-HWRYlI54NluY8U5eQ9TCeC5PhJa45GX2XouLN6UhzZTcJYQ7Ko6oaJw6PLT38LzBV4_Aw", "public_key": "bMUigy8O0jOnBxQ4Sc-5lwhIZ8LQVAhxMbR7qESVuUE", "signer_path": "lakehouse/data-extract/v1", "alg": "Ed25519", "signed": true } }, "written_at": "2026-09-04T06:39:59.832Z" }, "cited_facts": [ { "statement": "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.", "source": "SecurityWeek", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T10:38:13.000Z", "publisher_count": 1, "sources": [ "SecurityWeek" ] }, { "statement": "Greg Brockman stated that Astra meets the bar for artificial general intelligence, meaning it can match or exceed human capabilities across various tasks.", "source": "Decrypt", "instrument": "News", "claim_key": null, "published_at": "2026-09-03T19:24:24.000Z", "publisher_count": 1, "sources": [ "Decrypt" ] }, { "statement": "OpenAI released GPT-6 Astra on Thursday, which president Greg Brockman called a 'generational leap in capability' and the arrival of artificial general intelligence.", "source": "Decrypt", "instrument": "News", "claim_key": null, "published_at": "2026-09-03T19:24:24.000Z", "publisher_count": 1, "sources": [ "Decrypt" ] }, { "statement": "OpenAI rated its Astra model as its most dangerous model to date due to its ability to build exploits and bypass safety checks.", "source": "The Decoder", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T14:20:28.000Z", "publisher_count": 1, "sources": [ "The Decoder" ] }, { "statement": "OpenAI launched GPT-6 Astra, a new artificial intelligence model described by the company as the world's most intelligent and aligned.", "source": "9to5Google", "instrument": "News", "claim_key": null, "published_at": "2026-09-03T20:50:00.000Z", "publisher_count": 1, "sources": [ "9to5Google" ] }, { "statement": "In testing described by the company, Astra achieved a perfect score on ExploitBench, a benchmark that measures a model's ability to turn known vulnerabilities into working exploits.", "source": "SecurityWeek", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T10:38:13.000Z", "publisher_count": 1, "sources": [ "SecurityWeek" ] }, { "statement": "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.", "source": "SecurityWeek", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T10:38:13.000Z", "publisher_count": 1, "sources": [ "SecurityWeek" ] }, { "statement": "OpenAI stated that the 'Critical' classification requires additional safeguards before the model can be released.", "source": "SecurityWeek", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T10:38:13.000Z", "publisher_count": 1, "sources": [ "SecurityWeek" ] }, { "statement": "Astra built a browser exploit that escaped the sandbox and ran commands on the host during testing with security experts.", "source": "thenewstack.io", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T20:13:54.000Z", "publisher_count": 1, "sources": [ "thenewstack.io" ] }, { "statement": "The Critical cybersecurity threshold is reserved for models capable of finding vulnerabilities and developing exploits with far less human assistance.", "source": "thenewstack.io", "instrument": "News", "claim_key": null, "published_at": "2026-09-02T20:13:54.000Z", "publisher_count": 1, "sources": [ "thenewstack.io" ] }, { "statement": "GPT-6 Astra is the first model OpenAI has designated 'critical' under its Preparedness Framework for cybersecurity.", "source": "Decrypt", "instrument": "News", "claim_key": null, "published_at": "2026-09-03T19:24:24.000Z", "publisher_count": 1, "sources": [ "Decrypt" ] }, { "statement": "OpenAI acknowledged that Astra is the first system the company has rated capable of autonomously hacking well-protected systems without human guidance.", "source": "Decrypt", "instrument": "News", "claim_key": null, "published_at": "2026-09-03T19:24:24.000Z", "publisher_count": 1, "sources": [ "Decrypt" ] } ], "note": "A signature proves who filed this and that it has not changed since. It never makes a claim true." }