Research

Publication

• Apr 29, 2026
  Where the goblins came from
  How goblin outputs spread in AI models: timeline, root cause, and fixes behind personality-driven quirks in GPT-5 behavior.

• Apr 23, 2026
  GPT-5.5 System Card

• Mar 25, 2026
  Inside our approach to the Model Spec
  Learn how OpenAI’s Model Spec serves as a public framework for model behavior, balancing safety, user freedom, and accountability as AI systems advance.

• Mar 19, 2026
  How we monitor internal coding agents for misalignment
  How OpenAI uses chain-of-thought monitoring to study misalignment in internal coding agents—analyzing real-world deployments to detect risks and strengthen AI safety safeguards.

• Mar 10, 2026
  Improving instruction hierarchy in frontier LLMs
  IH-Challenge trains models to prioritize trusted instructions, improving instruction hierarchy, safety steerability, and resistance to prompt injection attacks.

• Mar 5, 2026
  Reasoning models struggle to control their chains of thought, and that’s good
  OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety safeguard.

• Mar 5, 2026
  GPT-5.4 Thinking System Card

• Mar 4, 2026
  Extending single-minus amplitudes to gravitons
  A new preprint extends single-minus amplitudes to gravitons, with GPT-5.2 Pro helping derive and verify nonzero graviton tree amplitudes in quantum gravity.

• Mar 3, 2026
  GPT-5.3 Instant System Card