🧰 GPT-Red

The items that name it 7

• Addendum to GPT-6 Luna System Card: October 28, 2026 Update
  paperas tools2026-10-29

  Automated red-teaming agent and adversarial training approach used to strengthen prompt injection resistance; its automated evaluations cover direct (instruction hierarchy) and indirect prompt injection, introduced in the GPT-5.6 system card

• Addendum to GPT-6 Astra System Card: GPT-6.1 Terra
  paperas classification.key entities2026-10-26
• GPT-6 Sol and GPT-6 Luna: October 2026 update
  articleas tools2026-10-07

  automated red-teaming agent approach using adversarial training to strengthen resistance to prompt injection while preserving legitimate task completion

• The state of AI for security: Measuring what matters most for building trust
  articleas tools2026-09-09

  OpenAI system using self-play red-teaming to find novel attacks that frontier models can't defend against

• GPT-6 Astra System Card
  writeupas tools2026-09-03

  Internal automated red-teaming attacker model/approach used to generate adversarial jailbreak and prompt-injection prompts and to adversarially train models for robustness

• AI Security Briefing
  articleas tools2026-07-18

  Automated AI system built by OpenAI to find security vulnerabilities (prompt injection attacks) in GPT models before release; trained via self-play reinforcement learning; succeeded in 84% of internal evaluation scenarios vs 13% for human red teamers; remains internal due to intentionally developed

• AI Security Briefing
  articleas classification.key entities2026-07-18

In every workspace you can read 12