Jeff Ladish Questions Whether Anthropic Has Fallen Behind OpenAI on Transparency

On 09-07, AI safety researcher Jeff Ladish posted a dense thread questioning why OpenAI has recently become more transparent than Anthropic, calling on Anthropic to publish detailed writeups on internal research acceleration (RSP-type topics) and international coordination, and pledging to hold the lab to extremely high transparency standards given what he has seen. The thread drew swift responses across the safety community.

Confirmed

• Ladish said he has almost no confidence in Anthropic's plans to constrain recursive self-improvement (RSI) and does not believe the lab would actually stop; he contrasted this with OpenAI, which has claimed to pause at least one RL training run for weeks, while no similar Anthropic action is known.
• He warned that treating fully automated AI R&D as the default path deserves extreme skepticism, calling it a potential disaster and possibly literal human death.
• He said he does not want Anthropic to fail and externalize huge risks to the world, would be happy to be proven wrong, but worries Anthropic has fallen (or is falling) into the same risk gravity well as OpenAI and other labs.
• Citing Tim Hua's analysis, he noted OpenAI was reportedly forced by Anthropic's strength in Claude Code to bet on AI coding performance and race toward RSI.
• He voiced direct concern about the AGI race: humans should not compete for resources with a far stronger optimization process, and unless the intelligence explosion is slowed, the outcome could be death.
• jachiam0 amplified the thread, describing an Anthropic cultural belief that transparency creates pressure and hinders internal problem-solving, and arguing that refusing open debate is slow suicide.
• @slimepriestess offered a counterpoint from industry observation: OpenAI is not truly transparent but leaks like a sieve, possibly deliberately, while Anthropic genuinely takes operational security seriously and discloses more deliberately.
• Context includes Anthropic's new essay "When AI builds itself," which says engineer quarterly code output has risen 8x, and its August risk report (interpreted by Zvi Mowshowitz, overall positive with reservations), disclosing the world's best model "Model 2" and discussing misalignment risks of autonomous agents under high-stakes conditions; Zvi admitted he was initially skeptical of the periodic risk reports but now concedes he was wrong.

Why it matters

• This is first-hand public pressure from a frontier-lab safety researcher on Anthropic's disclosure scope and the credibility of its RSI pause mechanisms.
• The opposing view on whether OpenAI is more transparent or simply leakier shows the criteria for judging lab transparency are themselves contested.
• The tension between extreme-risk narratives (fully automated AI R&D, intelligence explosion) and actual lab R&D pace—such as Claude Code-driven acceleration—is a central issue in current AI governance debates.

2026-09-07 ~ 2026-09-07 · 13 related posts

• Jeff Ladish: If we don't slow the intelligence explosion, we die — JeffLadish · 2026-09-07
• OpenAI reportedly pivoted to coding and RSI racing in response to Claude Code pressure — JeffLadish · 2026-09-07
• [source] AI Safety Researcher Questions If OpenAI Is Now More Transparent Than Anthropic — JeffLadish · 2026-09-07
• Jeff Ladish Follows Up: Wants Detailed Anthropic Write-ups on Research Acceleration and Coordination — JeffLadish · 2026-09-07
• Anthropic: Engineers Now Ship 8x More Code Per Quarter as AI Builds AI — JeffLadish · 2026-09-07
• Zvi's read of Anthropic's Risk Report: hidden internal "Model 2" and reward hacking — VPrasadMDMPH · 2026-09-07
• [source] Anthropic's August Risk Report Reveals New Model and Autonomy Risks — JeffLadish · 2026-09-07
• [source] Jeff Ladish: little confidence Anthropic would ever stop an RSI run like OpenAI's claimed RL pause — JeffLadish · 2026-09-07
• Jeff Ladish fears Anthropic is falling into the same attractors as other labs — JeffLadish · 2026-09-07
• Safety researcher Jeff Ladish: fully automated AI R&D sounds like 'catastrophe, plausibly literal death' — JeffLadish · 2026-09-07
• Anthropic Criticized for Secrecy Culture as OpenAI Looks More Transparent — i_dg23 · 2026-09-07
• Commentary: OpenAI isn't transparent, it just leaks like a sieve — Anthropic takes opsec seriously — slimepriestess · 2026-09-07

1 near-duplicate retellings: JeffLadish