Redwood Research
An independent AI safety research organization focused on adversarial robustness and interpretability research, dedicated to ensuring AI systems remain controllable under extreme inputs.
Core Facts
Timeline (last 90 days)
Independent research organizations METR and Redwood Research found multiple vulnerabilities while conducting a pilot review of OpenAI's internal monitoring system.
The pilot reviews by METR and Redwood Research are exploratory practice, and the applicable scope of their methodology and conclusions still requires joint discussion by the industry.
OpenAI said it has fixed the vulnerabilities affecting production safeguards, but researchers have not yet independently verified these fixes.
Within three weeks, METR and Redwood Research gained deep access to OpenAI's infrastructure, rather than merely reviewing documents or conducting interviews.
In the summer of 2026, OpenAI invited METR and Redwood Research to conduct a three-week embedded evaluation of its internal monitoring systems.
OpenAI stated that the findings of METR and Redwood Research helped it strengthen safeguards as it scaled its systems.
Redwood Research is a research organization focused on AI alignment and safety engineering, known for turning safety mechanisms into system components that can be engineered and verified.
AI safety research organization Redwood Research published an analysis article pointing out that latent-space reasoning architectures represented by COCONUT and full-bandwidth Transformers may greatly weaken the currently most effective AI supervision tool — chain-of-thought (CoT).
Redwood Research believes that coordinating to avoid latent-space architectures is more feasible than coordinating to maintain monitorability, because the architectural definition is relatively clear (operationalized as 'high opaque serial depth'), whereas monitorability evaluation requires a great deal of scientific judgment and leaves room for disagreement and motivated reasoning.
Redwood Research believes that interpretability techniques such as SAEs, probes, and NLAs will be difficult to sufficiently replace the supervisory function of CoT in the near term.
2 more timeline events
All Facts (12)
OpenAI announced that it will allow third parties to participate in safety evaluations during the training phase, covering the full pipeline of training, evaluation, and deployment, with partners including METR and Redwood Research.
65%UnverifiedIndependent research organizations METR and Redwood Research found multiple vulnerabilities while conducting a pilot review of OpenAI's internal monitoring system.
50%UnverifiedThe pilot reviews by METR and Redwood Research are exploratory practice, and the applicable scope of their methodology and conclusions still requires joint discussion by the industry.
50%UnverifiedOpenAI said it has fixed the vulnerabilities affecting production safeguards, but researchers have not yet independently verified these fixes.
50%UnverifiedWithin three weeks, METR and Redwood Research gained deep access to OpenAI's infrastructure, rather than merely reviewing documents or conducting interviews.
50%UnverifiedOpenAI stated that the findings of METR and Redwood Research helped it strengthen safeguards as it scaled its systems.
50%UnverifiedRedwood Research is a research organization focused on AI alignment and safety engineering, known for turning safety mechanisms into system components that can be engineered and verified.
50%UnverifiedRedwood Research believes that interpretability techniques such as SAEs, probes, and NLAs will be difficult to sufficiently replace the supervisory function of CoT in the near term.
50%UnverifiedAI safety research organization Redwood Research published an analysis article pointing out that latent-space reasoning architectures represented by COCONUT and full-bandwidth Transformers may greatly weaken the currently most effective AI supervision tool — chain-of-thought (CoT).
50%UnverifiedRedwood Research believes that coordinating to avoid latent-space architectures is more feasible than coordinating to maintain monitorability, because the architectural definition is relatively clear (operationalized as 'high opaque serial depth'), whereas monitorability evaluation requires a great deal of scientific judgment and leaves room for disagreement and motivated reasoning.
50%UnverifiedRedwood Research Chief Scientist Ryan Greenblatt noted that there is currently no good way to understand and supervise the activities and goals of AI collectives.
50%UnverifiedIn the summer of 2026, OpenAI invited METR and Redwood Research to conduct a three-week embedded evaluation of its internal monitoring systems.
50%