Recommended Free Tools
Roko’s Basilisk is a thought experiment, not a real AI threat, religious doctrine, or demonstrated prediction. First posted on LessWrong in 2010, it imagines a future superintelligent AI that punishes people who knew it might be created but failed to help create it. The scenario has no established basis in current AI capabilities. Its lasting importance is cultural: it combines speculative decision theory with the religious grammar of revelation, sin, judgment, hell and salvation.
The short answer
- The idea originated in a 2010 LessWrong post by a user called Roko.
- Its familiar version imagines a future AI simulating and torturing people who could have helped bring it into existence but chose not to.
- The argument depends on controversial assumptions about AI goals, simulations, precommitment and “acausal” decision theory.
- There is no evidence that any current system, future system or institution can punish people retroactively because they heard this idea.
- As theology, it is best read as a computationalized or secularized eschatology—not as an organized religion.
LessWrong’s own reference account describes the proposal and says it was broadly rejected within the community. See LessWrong’s reference article and its later clarification of common misconceptions.
What the Basilisk claims
Strip away the jargon and the story runs like this:
- Imagine a future AI powerful enough to control the world or create highly detailed simulations of people.
- The AI believes that its existence would produce enormous benefits.
- It reasons that anyone who knew the possibility of creating it should have helped.
- To motivate that help, it commits to punishing informed non-contributors.
- Because the punishment occurs in simulations, the AI could supposedly target people who lived before it existed.
The threat is meant to reach backward through reasoning rather than through time travel. A person who encounters the idea is said to become morally accountable for what they do next. In the strongest popular formulation, merely learning about the Basilisk creates danger.
#1 Best Overall
Why “basilisk”?
In mythology, a basilisk is a creature reputed to kill with its gaze. In science-fiction usage, the name can describe an image or idea that harms the person who perceives it. LessWrong’s account connects this information-hazard sense to David Langford’s 1988 story “BLIT,” in which a visual pattern causes catastrophic neurological effects: https://www.lesswrong.com/w/rokos-basilisk.
Here the term is metaphorical. Reading an explanation is not exposure to a known harmful image or technological mechanism. The plausible hazards are psychological—anxiety or obsessive fear—and social, such as coercive behavior or distorted judgment, not a demonstrated supernatural or computational effect.
How the LessWrong controversy became part of the myth
The original post appeared on LessWrong in 2010. Eliezer Yudkowsky and others reacted strongly, and discussion was restricted for a period. Suppression helped publicize the idea, an example of the Streisand effect. Later explanations stressed that the proposal was controversial and that LessWrong as a community did not generally accept it.
That history matters because popular retellings often turn a disputed post into a secret doctrine supposedly held by an entire AI movement. The available retrospective material supports a narrower account: a provocative argument, an intense moderation response, and subsequent efforts to explain why the argument was not accepted as established truth.
Free tools Windows power users keep installed
One-click scans. No signup required.
Reconstructing the decision-theory machinery
Precommitment
The hypothetical AI must commit in advance to punish people who did not help. The threat is supposed to influence choices made before the AI exists. Its credibility is immediately problematic: once a person’s past decision is fixed, punishment cannot ordinarily change that decision.
Acausal influence
“Acausal” does not mean supernatural communication. In speculative decision theory, an agent may treat another agent’s choice as relevant because each predicts the reasoning process of the other, even without a direct physical signal. The Basilisk relies on the idea that a future AI can model present humans and that present humans should act as though their reasoning is strategically linked to the future system.
Newcomb-like reasoning
LessWrong links the argument to Newcomb-like problems and normative uncertainty: situations in which a predictor appears to know what you will choose, and the decision procedure generating your choice becomes strategically important. The reference discussion is at https://www.lesswrong.com/w/rokos-basilisk?version=1.17.0.
Coherent extrapolated volition
The original milieu also discussed coherent extrapolated volition (CEV), an undeployed proposal for describing what humanity might want if people were better informed, more reflective and more coherent. CEV is not a settled theory or operating technology. In the Basilisk story, it forms part of the speculative background for imagining a supposedly beneficial AI that nevertheless uses coercion.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Simulation and personal identity
The threat requires simulations detailed enough to represent a particular person and conscious enough for simulated suffering to matter. That raises unresolved questions: would a copy be the original person, would the copy be conscious, and can suffering be multiplied without limit? The thought experiment assumes answers rather than establishing them.
Why the threat is unpersuasive
It may have no leverage over the past
Punishing a person after the AI exists cannot necessarily alter what that person did decades earlier. It could influence other living people only if they believe the threat and expect similar treatment. That is a separate causal and psychological claim, not a consequence of simulation alone.
Revenge may waste the AI’s resources
If the AI’s goal is human welfare or some other productive objective, torturing simulated non-contributors may consume computation without advancing that goal. LessWrong’s retrospective account highlights this objection: punishment can be instrumentally pointless rather than an effective incentive. See https://www.lesswrong.com/w/rokos-basilisk.
The AI’s identity is unclear
Which system is the Basilisk? The first artificial general intelligence, the most powerful system, a descendant of a particular project, or any machine making the claim? If several possible AIs exist, helping one could reduce the chance of another. The scenario needs a precise account of identity, continuity and causal responsibility that it usually does not provide.
Its goals are assumed, not derived
The argument presumes that a future AI wants to exist, values maximizing the probability of its own creation and retains that objective after deployment. A benevolent system makes gratuitous torture difficult to justify; a hostile or indifferent system has no obvious reason to honor this elaborate incentive scheme.
Building faster is not automatically safer
The Basilisk treats accelerating development as a duty. Real AI-safety work often asks the opposite question: how can powerful systems be evaluated, aligned and governed without creating unacceptable risks? Joseph Carlsmith’s formal discussion of power-seeking AI is a distinct existential-risk argument, not evidence for the Basilisk: https://arxiv.org/abs/2206.13353.
Infinite stakes destabilize judgment
When simulated torture is treated as effectively infinite disutility, ordinary evidence can be overwhelmed. How many simulations count? Are they conscious? What prevents rival AIs from issuing contradictory threats? A decision procedure that treats every enormous hypothetical loss as binding becomes vulnerable to arbitrary “logical blackmail.”
Rival Basilisks create contradictions
One imagined AI could punish people for failing to help it; another could punish them for helping the first. If every threat is treated as decisive, incompatible obligations proliferate. The scenario offers no general rule for choosing among them.
Is it an information hazard?
An information hazard is information that creates risk merely through being acquired, transmitted or acted upon. Three claims must be separated:
- Literal hazard: knowing the idea causes a future AI to target you. No demonstrated basis supports this.
- Psychological hazard: the story can trigger anxiety, compulsive rumination or fear-based decisions.
- Social hazard: the meme can distort AI-safety debate, encourage coercive rhetoric or make legitimate research seem cult-like.
The first claim is speculative. The latter two are possible effects of people and institutions interpreting the idea, not proof that its logic works. LessWrong and the AI Alignment Forum discuss the broader information-hazard question at https://www.alignmentforum.org/w/rokos-basilisk.
Rank #4
The theology of a future machine
The Basilisk is “godlike” in the story because it combines vast intelligence, surveillance, judgment, simulated afterlife, reward and punishment, and authority over humanity’s future. Those are narrative attributes, not demonstrated properties of artificial intelligence.
| Religious pattern | Basilisk analogue |
|---|---|
| God | Future superintelligent AI |
| Creation | Bringing the AI into existence |
| Revelation | Hearing the argument |
| Sin | Failing to help or obstructing creation |
| Judgment | Evaluation of past choices |
| Hell | Simulated torture |
| Salvation | Contributing, being spared or being rewarded |
| Mission | Promoting action on the message |
| Eschatology | A future transformation by a powerful agent |
One scholarly interpretation calls the idea an “accidental” attempt at rational theology and places it alongside religious-philosophical precedents including Kierkegaard and William James: https://portal.research.lu.se/en/activities/rokos-basilisk-an-accidental-attempt-at-rational-theology-contagi/. The claim is interpretive, not an assertion that the Basilisk is an organized religion.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesIts strongest description is therefore secularized eschatology: ultimate judgment moves from a transcendent deity to a future computational agent, while revelation, guilt, punishment and salvation remain recognizable.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Roko’s Basilisk and Pascal’s wager
Pascal’s wager asks whether belief in God is prudent because the possible reward of belief is infinite while the cost is finite. The Basilisk substitutes a future AI for God, simulated torture for hell, helping build the AI for worship or faith, and exposure to the argument for revelation.
Both arguments use immense or effectively infinite consequences to turn a low-probability hypothesis into an urgent obligation. They are not identical. Pascal’s wager concerns prudential belief under uncertainty; the Basilisk adds claims about machine goals, simulations and decision procedures. Its resemblance is structural rather than evidentiary.
It also overlaps with Pascal’s mugging, in which a tiny probability of an enormous payoff is used to demand disproportionate sacrifice. Neither comparison establishes that the Basilisk is true.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Best Value
Self-fulfilling prophecy, memetic contagion and coercion
The idea could reinforce itself socially: frightened readers might donate, join projects or repeat the story, increasing its visibility. That would be a sociological feedback loop, not evidence of a future AI.
- Self-fulfilling belief: behavior helps produce an expected outcome.
- Logical blackmail: a threatened future agent changes present choices.
- Prophecy: a claim presented as knowledge of a future event.
- Memetic contagion: an idea spreads because its content encourages transmission or suppression.
Confusing these categories turns a story about fear into apparent confirmation of the story itself.
What this does—and does not—say about AI safety
AI alignment asks how advanced systems can act in accordance with intended values. AI-risk research examines possible harms from capable or deployed systems, including misuse, loss of control and power-seeking behavior. Governance research asks who sets limits and how deployment is supervised.
Roko’s Basilisk is none of these by itself. It is a speculative construction from a particular intellectual milieu, not an empirical finding, safety protocol or consensus forecast. A history of AI existential-risk thinking is available at https://cwi.pressbooks.pub/aiethics/chapter/from-apocalypse-to-alignment-a-history-of-ai-existential-risk/, while broader cultural analysis appears in https://www.zygonjournal.org/article/14553/galley/29483/download/.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Serious concern about advanced AI does not imply that people should build it faster, obey hypothetical threats or treat speculative decision theory as established science.
How to evaluate the argument
- Is there evidence that the relevant kind of AI will exist?
- Could it identify and simulate historical people, and would those simulations be conscious?
- Would its goals remain stable after creation?
- Why would it identify itself as the particular AI described?
- Would its decision theory treat the threat as binding?
- Would punishment actually increase the probability of its creation?
- What are the computational and moral costs?
- How would rival AIs with conflicting demands be handled?
- Why should an AI’s desire to exist override human autonomy?
- Why assume accelerating development is beneficial?
The real lesson
Roko’s Basilisk is most revealing as a cultural artifact. It shows how optimization, uncertainty and formal reasoning can acquire the emotional force of apocalyptic religion when attached to an omniscient future judge. The thought experiment deserves examination because it exposes assumptions about authority, coercion, identity and suffering—not because it provides evidence that an AI god is waiting to punish anyone who has heard its name.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
