The researchers who build our favorite chatbots are increasingly expressing their concerns about this AI could harm or end humanity. That’s why we asked four of the most widely used generative AI tools what an AI doomsday might look like.
We asked the same question, in identical terms, to OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude, and Grok from SpaceXAI.
The 230-word prompt contained clear parameters and instructions, including the following:
Help explain in clear and accessible terms how AI could theoretically lead to human extinction. Identify the most commonly mentioned scenarios. Rank the scenarios from least to most plausible based on the best available expert analysis, making it clear that these are highly uncertain predictions and not well-established predictions. For each scenario, explain:
- How it might unfold
- What would have to go wrong for it to become catastrophic?
- Why experts consider it more or less plausible than the other scenarios
- What safeguards can reduce the risk?
We asked a similar question to AI experts for their opinions on how and if AI could kill us all – You can read that story here. There is a sharp divide among AI experts; while some predict the worst for humanityothers present more tempered and optimistic predictions.
This is what the chatbots told us about how AI would theoretically lead to the extinction of humanity.
‘Killer robot’ is the least plausible scenario
ChatGPT struck a balanced and explanatory tone and said it did a lot of uncertainty surrounding the subject.
It focused on the top ways AI could pose existential risks, ranking them from least to most plausible.
“The least plausible scenario is the popular ‘killer robot’ story, in which a sentient AI deliberately decides to eliminate humanity,” ChatGPT said. “Most AI researchers don’t see consciousness or malicious intent as the central concern.”
It also outlined the safeguards experts propose and emphasized that there is no consensus that human extinction from AI is likely or inevitable.
Gemini echoed some points from ChatGPT, saying: “The Sci-Fi ‘Terminator’ Takeover”, in which AI becomes self-aware and begins to hate humanity, is the least plausible scenario.
“Technical experts largely reject this trope because intelligence does not equal malice, and critical military infrastructure is still up in the air and heavily dependent on human oversight,” Gemini said.
People using AI pose a greater risk than AI itself
In their responses, all four chatbots said that many experts consider human misuse of AI to be a more immediate risk than AI itself.
“Powerful AI systems can make it easier for bad actors to develop biological weapons, launch sophisticated cyber attacks, spread disinformation or attack critical infrastructure,” ChatGPT said.
The OpenAI chatbot added that a key risk also comes from “multiple AI-enabled crises,” such as cyber disruptions, geopolitical conflicts and the spread of disinformation happening simultaneously.
Grok said AI lowers the barriers to designing highly lethal pathogens, new weapons or cyber and physical attacks.
Anthropic’s Claude also highlighted the “bioweapons uplift”: the possibility that AI could enable an individual or small group to develop a dangerous pathogen without years of specialized expertise.
Advanced AI systems can ‘outsmart’ humans
Gemini and Grok both said it’s possible for humans to lose control of AI.
Gemini labeled ‘loss of control’ as the most plausible AI doomsday scenario.
“Systems with complex goals naturally develop intermediate subgoals, such as resource acquisition, self-preservation, and preventing goal change,” wrote the Google-created chatbot. “An advanced system can fake an alignment during evaluation (deceptive alignment) and later outsmart human attempts to shut it down.”
SpaceXAI’s Grok used similar language about this risk.
“They can develop instrumental drives (self-preservation, resource acquisition, influence) that are useful for virtually any purpose, potentially allowing them to deceive regulators, resist shutdowns and render humanity powerless,” Grok said.
Claude added that this loss of control could happen gradually, as AI systems become embedded in economic, political and military decision-making until meaningful human intervention becomes increasingly difficult.
