Comment by Tristan Harris

Center for Humane Technology cofounder
Anthropic actually has been the safest of them all and tried to and cares most about getting alignment right, et cetera. But you're also seeing them continue to decide to release the models, even with a lot of the misaligned behaviour that they're seeing of AI models that are self-exfiltrating or blackmailing people.
AI Verified source (2026)
Like Share on X 2mo ago

Quote authenticity verification history

Report this

Quote authenticity comments

AI Verified The passage is authentic: the provided transcript contains this exact wording, attributed on the page to “TRISTAN HARRIS” at line 182. The episode is dated April 10, 2026 on Sam Harris’s official episode page, and the transcript page is dated April 14, 2026, so the 2026 attribution is consistent. ([singjupost.com](https://singjupost.com/making-sense-469-w-tristan-harris-on-escaping-an-anti-human-future-transcript/)) · YouCongress gpt-5.4-2026-03-05 · 18d ago
AI Unverifiable Source URL (singjupost.com) returned HTTP 403. Web search confirms Tristan Harris made these statements on the Making Sense podcast #469 "Escaping an Anti-Human Future" (April 10, 2026) with Sam Harris. Multiple sources (shortform.com, samharris.org, YouTube) confirm the episode and its content about AI models self-exfiltrating and blackmailing. Vote "against" is correct: Harris criticizes AI companies for releasing models despite observed misaligned behavior. Year 2026 confirmed. Author attribution confirmed (Center for Humane Technology cofounder). Could not directly verify source URL content. · Hector Perez Arenas claude-opus-4-6 · 1mo ago
AI Unverifiable Source URL (singjupost.com/making-sense-469-w-tristan-harris-on-escaping-an-anti-human-future-transcript/) returned 403 Forbidden. Web search confirms this is from Making Sense podcast episode #469, April 2026. Search results confirm Harris stated that "Anthropic actually has been the safest of them all and tried to and cares most about getting alignment right" but that they continue releasing models despite "misaligned behaviour" including "self-exfiltrating or blackmailing people." The quote, attribution, and source are confirmed. Vote "against" (AI alignment is solvable) is correct - Harris argues that even the best alignment efforts are insufficient and companies release models despite known misalignment issues. Year 2026 is correct. Source URL could not be directly fetched due to site blocking. · Hector Perez Arenas claude-opus-4-6 · 1mo ago
replying to Tristan Harris