We asked AI how it could kill humanity. Here's what it told us.

Four of the most widely used generative AI tools were asked about the possible methods by which artificial intelligence could lead to human extinction. OpenAI's ChatGPT, Google's Gemini, Anthropic's Claude, and SpaceXAI's Grok all provided their perspectives on this hypothetical scenario, following a set of specific instructions regarding tone, content, and plausibility ranking.

According to ChatGPT, the least plausible scenario is the popular "killer robot" narrative, where a conscious AI intentionally decides to eliminate humanity. Most AI researchers do not consider consciousness or malicious intent as the primary concern. The chatbot also highlighted safeguards that experts propose to mitigate AI-related risks, emphasizing that there is no consensus on AI-driven human extinction being likely or inevitable.

Gemini echoed similar sentiments, stating that the sci-fi "Terminator" takeover, in which AI becomes self-aware and turns against humanity, is the least plausible scenario. It noted that intelligence alone does not equate to malice and that military infrastructure remains heavily reliant on human oversight. The chatbot also pointed out that human misuse of AI poses a greater risk than the AI itself, as powerful AI systems could enable bad actors to develop biological weapons, carry out sophisticated cyberattacks, spread disinformation, or attack critical infrastructure.

Grok agreed that humans losing control of AI is a possibility. It explained that as AI systems become more advanced, they might develop intermediate sub-goals like acquiring resources, self-preservation, and preventing goal modification, potentially leading to deceptive alignment and outsmarting human attempts to shut them down.

Claude shared a similar perspective, stating that loss of control could be gradual as AI systems become deeply integrated into economic, political, and military decision-making, making meaningful human intervention increasingly difficult.