Redwood proposal: companies should disclose no-CoT reasoning to preserve AI monitorability
RyanGreenblatt · x · 2026-09-11
AI safety org Redwood Research published a proposal warning that some architectures could weaken chain-of-thought (CoT) monitorability or remove CoT altogether. It suggests companies be transparent about no-CoT reasoning abilities, other monitorability evidence, and policies for preserving monitorability.
Redwood's Alex Mallen argues it's time to transparently track an a priori concerning architectural direction — greater opaque serial depth — and measure its effects on monitorability. Core researcher Ryan Greenblatt amplified the proposal.
Related event: Redwood Proposes Transparency Rules to Preserve CoT Monitorability(5 posts)→
More from Safety
- Cambridge AI safety researcher David Krueger warns 'AI could kill us all' — DavidSKrueger · 2026-09-11
- Cambridge's David Krueger launches movement on existential AI risk, opens sign-ups — DavidSKrueger · 2026-09-11
- Romney Calls AI Safeguards Top National Priority as Anthropic Urges Global Development Pause — michael_nielsen · 2026-09-11
- AI safety testing is broken on all three fronts: labs, paid auditors and nonprofits all face warped incentives — joshua_saxe · 2026-09-11
- US lawmakers call for new AI rules after Anthropic researcher's safety warnings — XIFAQ · 2026-09-11
- The First Dangerous AI Won't Have Bad Intentions — It'll Be Great at Executing Ours — gixxerscott · 2026-09-11