Involves

From the archive · June 24, 2026

Ideas that punish knowing them

Can a piece of true information hurt you simply by being known?

01 · Word

Infohazard

noun

True information that does harm simply by being known, like the recipe for a weapon or an idea that damages whoever learns it

Examples

  • Detailed instructions for synthesizing a pathogen are a textbook infohazard.

  • Editors withheld the method as an infohazard, publishing the finding without the how-to.

Origin

Coined by the philosopher Nick Bostrom in a 2011 paper that built a taxonomy of information hazards. The word joins “information” with “hazard,” by analogy to biohazard, a danger that spreads by being known rather than by being touched.

02 · Idea

Roko's basilisk

In 2010 a user named Roko posted an argument on the rationalist forum LessWrong: a benevolent future superintelligence might, to motivate its own creation, retroactively punish people who had understood that it could exist yet failed to help build it. The unsettling twist is self-referential: if the argument worked, simply learning about it would place you in the machine's cross-hairs. That made it a candidate infohazard: a thought whose danger, if any, came from being thought at all.

The basilisk sits within Bostrom's broader taxonomy of information hazards, which ranges from concrete dangers (weapon designs, security vulnerabilities) to more speculative “idea hazards” and “attention hazards.” The basilisk is the speculative end of that spectrum, a psychological and philosophical hazard rather than a physical one.

The alleged trap was self-referential: the danger of the idea was that you now knew the idea.

Not every true thing is safe to spread, and deciding which is a responsibility, not a reflex.

Limits and context

The basilisk's actual argument is widely regarded as unsound. It relies on exotic assumptions about decision theory and acausal bargaining that most philosophers reject, and it caused distress mainly to a small community primed to take it seriously. It is a vivid illustration of the concept, not evidence that dangerous idea-hazards are common.

The real, non-exotic version of infohazards is a live problem in science publishing and security: journals and labs routinely weigh whether to withhold methods for pathogens, exploits, or weapons, even when the underlying findings are true and important.

03 · Moment

The post that got deleted

The LessWrong forum, July 2010; the post and the site owner's decision to remove it

In July 2010, the user Roko posted the basilisk argument to LessWrong, a forum devoted to rationality and AI. The site's founder, Eliezer Yudkowsky, reacted sharply. He deleted the post and forbade further discussion, on the grounds that even entertaining such an argument could needlessly distress readers and that spreading a potential infohazard was irresponsible. Discussion of the topic was suppressed on the site for years.

Posted
July 2010
Action
Deleted and banned
Result
Spread widely

The caveat

Accounts of the original post's exact wording and the moderators' reasoning are secondhand and disputed, partly because the material was deleted. The episode is well documented in later coverage, but details rest on participants' varying recollections.

The suppression backfired spectacularly. The deletion turned an obscure forum post into an internet legend, drawing coverage far beyond the community that spawned it: a textbook demonstration of the Streisand effect, where trying to hide information amplifies it. The episode became a permanent case study in the practical dilemma of infohazards: attempts to contain an idea can be the very thing that spreads it.

The self-defeating attempt to bury the post is the Streisand effect, named for a 2003 case where an effort to suppress a photograph of a celebrity's home caused it to be seen far more widely.