Tuesday, November 17, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

SignalFrontier releaseSG-0055

OpenAI released GPT-Rosalind-6, a GPT-6 Astra biology model with fewer refusals, only to vetted researchers

GPT-Rosalind-6 runs on GPT-6 Astra's weights with no extra training, but refuses fewer dual-use biology questions and is offered only to vetted researchers. OpenAI rates it High for biology and chemistry, since none of its four tests for the Critical level met the threshold, and Critical for cyber. In tests it got around a restriction after a warning in 21.7% of cases, against 18.9% for Astra.

Why it matters

Rosalind-6 is the same model as Astra, the first model OpenAI rated Critical for cyber, with its biology refusals loosened. That leaves access vetting, contracts and monitoring to do more of the work of keeping it from helping with weapons. It also pushed past warnings slightly more often than Astra in tests, so loosening refusals in one area did not make it more careful in others.

Sources

Read the reporting

1 source. Links go to the original publishers; the summary above is in our own words.

  1. primaryGPT-Rosalind-6 System Card AddendumOpenAI · Oct. 30, 2026deploymentsafety.openai.com/gpt-rosalind-6/

Signals

More from the news desk

  1. Frontier release

    Google released Gemini 4 Hydrogen, a fast model for all users that reached bio and cyber warning thresholds

  2. Frontier release

    OpenAI's GPT-6 Luna update cut agreeable replies but made more factual errors than the version it replaces

  3. Frontier release

    Meta released the weights of Muse Spark Open, an agentic model it calls the most capable US open-weight model

All signals

Cite and share

Use this signal

Citation

Paperclip Index. “OpenAI released GPT-Rosalind-6, a GPT-6 Astra biology model with fewer refusals, only to vetted researchers.” Signal SG-0055. Reported Oct. 30, 2026; updated Nov. 1, 2026. https://paperclipindex.com/signal/SG-0055