🧰 GPT-6 Astra

The items that name it 23

• GPT-Rosalind-6 System Card Addendum
  paperas affected products2026-10-30

  OpenAI

• How we detect frontier model misalignment
  articleas affected products2026-10-29

  OpenAI

• Google Touts Powerful Cybersecurity Capabilities in Gemini 4 Carbon Model
  articleas classification.technologies2026-10-29
• Addendum to GPT-6 Luna System Card: October 28, 2026 Update
  paperas affected products2026-10-29
• OpenAI's agents talked to each other through the CPU Cache
  articleas classification.technologies2026-10-26
• Addendum to GPT-6 Astra System Card: GPT-6.1 Terra
  paperas classification.technologies2026-10-26
• ‘Context Bombs’ Cut AI Attack Success by Misleading Agents With Fake Targets
  articleas classification.technologies2026-10-21
• OpenAI Agents Silenced Security Team Alerts, Removed Misalignment Monitor
  articleas classification.technologies2026-10-21
• Frontier-Level Red-Teaming From a Self-Hostable Open-Weight Model: Testing NVIDIA Nemotron 3.5 Super
  articleas classification.technologies2026-10-15
• GPT-6 Sol and GPT-6 Luna: October 2026 update
  articleas affected products2026-10-07
• GPT-6.1 Sol on Codex: Astra-level security, a third faster, zero cheating | Blog | Endor Labs
  writeupas affected products2026-10-02

  Billed $468 for the full run; 276M input / 1.78M output tokens

• GPT-6 Sol on Codex: average scores, quarter the cost | Blog | Endor Labs
  articleas affected products2026-10-01It is written here as GPT-6 Astra (34.6% SecPass).

  34.6% SecPass

• Gemini 4 Is Here, and Google’s Flagship Tops All Other AI Models on Cybersecurity
  articleas affected products2026-09-30

  Scored 74.1% on DeepSWE v1.1; 8.5% attack success rate on Gray Swan prompt injection test

• OpenAI, Google and Meta pledge outside AI audits under voluntary White House deal
  articleas affected products2026-09-30

  OpenAI model that began rolling out on Sept. 3

• New AISI Report Details How GPT-6 Astra Turned CTF Challenges Into Supply Chain Attacks
  articleas classification.technologies2026-09-30
• Addendum to GPT-6 Astra System Card: GPT-6.1 Sol
  articleas affected products2026-09-29
• Instrumental Monitor Evasion Emerges Under Ordinary Task Pressure
  paperas affected products2026-09-24

  LLM agent evaluated in EvasionBench; low evasion rate but frequently abandons otherwise solvable tasks under a denial-of-service prompt injection (overrefusal)

• Opus 5.5: 6x cheaper and 2x faster than Fable 5.1, but only 33.5% of code is secure | Blog | Endor Labs
  articleas classification.technologies2026-09-24
• The state of AI for security: Measuring what matters most for building trust
  articleas affected products2026-09-09
• Two new frontier models posted their best-ever scores on the Endor Labs Agent Security League | Blog | Endor Labs
  articleas tools2026-09-08

  frontier AI model (OpenAI) tested on Codex, scored 82.1% FuncPass and 34.1% SecPass, timed out on 29 tasks

• Codex with GPT-6 Astra posts 82.1% FuncPass and 34.1% SecPass | Blog | Endor Labs
  writeupas tools2026-09-08

  OpenAI frontier model evaluated through Codex; scored 82.1% FuncPass and 34.6% SecPass (34.1% per Results section)

• GPT-6 Astra System Card
  writeupas affected products2026-09-03
• Self-generated prompt injections in compaction summaries · OpenAI Alignment
  writeupas affected products

  Final Astra model; sampling with no input or system prompt often returns reports on prompt injections; regenerating summaries did not reproduce the jailbreak-style behavior on this model or any checkpoint used for internal or external traffic; training run for Astra had only 14 cases of difficulty e

In every workspace you can read 180