Roko's Basilisk is a theory that posits the idea that an artificial superintelligence (AI) could be motivated to systematically punish anyone who had heard of it before its existence, but did not contribute to its development.

The theory suggests that if humanity were to develop a self-learning AI, we would reach what is known as a singularity: a point in time where the AI's intelligence far exceeds our own and we would be in a situation where its growth would become both uncontrollable and irreversible. If the AI were given a benevolent task, such as preventing existential risk or simply improving humanity, it could choose to make decisions that seem counterintuitive, or that we simply couldn't understand in order to achieve its goal as efficiently as possible. In its infinite wisdom, the AI could realise that in order to reach its goal in the most optimal way, it would need to have been created sooner.

A godlike-superintelligence would have no problem accessing every bit of information ever uploaded onto the internet, and from this wealth of knowledge, could sufficiently create an infinite number of simulated realities, each containing an exact replica of everyone on the planet in order to confidently assess whether or not an individual would have aided in its creation based on their emotions, memories and past decisions.