Artificial Intelligence
Redwood Research
A technical AI safety research organization focused on 'AI control' -- techniques to ensure AI systems cannot subvert safety measures even if misaligned -- advising Google DeepMind and Anthropic on assessing and mitigating risks from misaligned AI agents. Redwood Research partnered with the UK AI Security Institute to publish 'A sketch of an AI control safety case,' describing how developers can construct a structured argument that models are incapable of subverting control measures, and hosted ControlConf 2026, concluding that control techniques are now load-bearing for the safety of real-world agent deployments.
Sources
Every claim traced to a primary source — evidence, recent activity, and insider filing behavior, all in one place.
Evidence
1 source · 48% avg. source confidence
- Redwood Research — Redwood Research and UK AISI publish AI control safety case, host ControlConf 202648% source confidence
Dependency Map
What Redwood Research depends on below, and who depends on Redwood Research above — click any node to make it the new center, 3 levels deep.
Connection type
Click any node to make it the new center. Scroll to zoom, drag to pan.
The Story So Far (last 6 months)
Auto-generated from this entity's dated milestones, relationship updates, and sourced evidence — not AI-written, just sorted.
Redwood Research's Timeline
A sourced, dated history of Redwood Research's key moments — founding to present.
Jan 2021 · Founded
Redwood Research was founded in Berkeley, California to conduct technical AI alignment and control research.
Jan 2026 · Publishes AI control safety case with UK AI Security Institute
Redwood Research partnered with the UK AI Security Institute to produce 'A sketch of an AI control safety case,' describing how AI developers can construct a structured argument that their models are incapable of subverting control measures even if misaligned.
Apr 2026 · Hosts ControlConf 2026
Redwood Research hosted ControlConf 2026, including a workshop on AI futurism and threat modeling, concluding that control techniques have become load-bearing for the safety of real agent deployments as AI companies begin building initial control measures.