Ryan Greenblatt on AI alignment is solvable

Ryan Greenblatt

Chief Scientist at Redwood Research; AI safety researcher; lead author of "Alignment faking in large language models"

We don't know how AIs are aligned. A somewhat crazy aspect of the current situation is that we have very little confirmed public information about why frontier AIs end up being apparently behaviorally aligned. And more generally, we don't know what factors in training are most relevant for behavioral alignment. [...] AI companies should be much more transparent about this.

AI Unverifiable source (May 8, 2026)

4mo ago

Policy proposals and claims

AI alignment is solvable

votes Against

Statement relation verification history Unverified Report this

No statement relation verification comments yet.

Vote inference verification history Unverified Report this

No vote answer verification comments yet.

Quote authenticity verification history

Report this

AI Unverifiable The quote text is strongly corroborated: Ryan Greenblatt’s AI Alignment Forum shortform entry contains the full wording verbatim, and a TwStalker search result shows the same full text attributed to Ryan Greenblatt @RyanPGreenblatt as an X post, so the [...] omission is faithful. But the supplied X URL itself returned no readable content in this environment, so I cannot directly confirm that exact source page contains it. The status ID is at least consistent with a 2026-05-08 UTC post date. ([alignmentforum.org](https://www.alignmentforum.org/posts/FG54euEAesRkSZuJN?commentId=XjLhEzNgHsWBQE3p7)) · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Disputed The passage is genuinely attributable to Ryan Greenblatt: a TwStalker mirror shows a post under @RyanPGreenblatt with this text, and Greenblatt’s AI Alignment Forum shortform contains the same wording. However, the supplied quote is not strictly verbatim, because the source says “most relevant for (behavioral) alignment,” not “most relevant for behavioral alignment.” I could not directly verify the X page itself, but another reliable source contains the passage. ([twstalker.com](https://twstalker.com/JoinTorchbearer?utm_source=openai)) · YouCongress gpt-5.4-2026-03-05 · 4mo ago

AI Unverifiable Source URL (x.com) returned HTTP 403. Web search confirms Ryan Greenblatt posted this exact quote on X on May 8, 2026. The search result title contains the quote verbatim. Multiple LessWrong/Alignment Forum posts reference the same statements. Vote "against" is correct: Greenblatt highlights that we don't know how AIs are aligned or what factors in training drive alignment, calling for more transparency. Year 2026 confirmed. Author attribution confirmed (Chief Scientist at Redwood Research). Could not directly verify source URL content. · Hector Perez Arenas claude-opus-4-6 · 4mo ago

AI Unverifiable Source URL (x.com/RyanPGreenblatt/status/2052803011915980856) returned 403 Forbidden. Web search confirms the exact tweet from Ryan Greenblatt posted May 8, 2026: "We don't know how AIs are aligned. A somewhat crazy aspect of the current situation is that we have very little confirmed public information about why frontier AIs end up being apparently behaviorally aligned." The quote, attribution, and source are confirmed. Vote "against" (AI alignment is solvable) is correct - Greenblatt highlights that we don't understand how alignment works, implying it's not yet truly solved. Year 2026 is correct. Source URL could not be directly fetched due to X blocking. · Hector Perez Arenas claude-opus-4-6 · 4mo ago

replying to Ryan Greenblatt

Other opinions from Ryan Greenblatt

See all

Ryan Greenblatt votes For

We should adopt AI 2040 Plan A

Plan A seems like a good plan for handling powerful AI, or at least the best plan anyone’s written up.

AI Verified source (Jul 11, 2026)

added 1mo ago

Ryan Greenblatt votes For

The U.S. and China should establish Mutually Assured Compute Destruction

Plan A seems like a good plan for handling powerful AI, or at least the best plan anyone’s written up. Many choices initially seem crazy, but are actually pretty carefully considered. Plan A isn’t likely to happen, but pushing for something like this...

more AI Verified source (Jul 11, 2026)

added 2mo ago

Ryan Greenblatt votes For

Mandate third-party audits for major AI systems

AI companies should get third-party security experts to assess (and possibly also red-team/pen-test) their security against key threat models and then publish the high-level findings of this assessment.

AI Verified source (Apr 27, 2026)

added 2mo ago