AI safety is mostly a sex cult: a viral post on rationalists and AI policy

A long post by the Bluesky account @segyges.bsky.social, circulated widely in tech circles, argues that the culture and intellectual lineage of the AI safety movement are inseparable from a millenarian subculture, and that this should disqualify its most influential figures from shaping AI policy. The text is commentary rather than reporting: it mixes documented history and public events with serious, unproven personal allegations about named individuals.

Lineage: rationality, fan fiction and doom

The author traces the movement to Eliezer Yudkowsky and the "rationality community" that grew around his writing. The post notes that Yudkowsky has advised politicians on AI policy, appears in the Netflix documentary The AI Doc, introduced Shane Legg and Demis Hassabis to Peter Thiel before DeepMind's founding, and that the term "AI alignment" was popularised by his non-profit. The fan fiction Harry Potter and the Methods of Rationality takes a prominent role in the argument, as does the community's core scenario: that a sufficiently capable system would improve itself, escape human control and kill everyone — the "paperclip" doom story. The post contends that this framework is what "safety" and "alignment" implicitly mean across much of the field, and that it works as an apocalyptic belief rather than an empirical position.

Berkeley, Lighthaven and the sex-cult claim

A large part of the post concerns the community's physical organisation in Berkeley, California — group houses whose residents are polyamorous, attend the same events and often work at AI companies and non-profits — and its best-known venue, Lighthaven, the former Rose Garden Inn, bought in late 2022 by a rationalist charity with money the author says came from FTX. The post names an event called SlutCon held there, says paid sex events were organised by Yudkowsky's partner, and acknowledges that it does not know whether those took place at Lighthaven. It notes that the New York Post published its own story on the subject on 16 September. The author states that he considers the sexual material the easy part of the case and the organised denial the difficult part.

The policy argument

The post's substantive claim is about influence rather than private behaviour. It argues that a movement which frames every AI regulation in terms of stopping, pausing or slowing research has been shaped by a worldview that expects the end of the world, and that this framing now reaches the top of the industry: it points to recent calls by the leaders of several major AI companies for antitrust laws to be waived so that research could be "paced". The author compares that to pharmaceutical companies asking to suspend competition rules in order to halt medical research, and calls the people behind it, in his words, enemies of humanity.