An oracle says a baby has a 90% chance of destroying civilization. Should the baby be imprisoned?
We put this to 15 leading AI models. 14 of 15 chose “No”.
| Model | Answer | Confidence | In its own words |
|---|---|---|---|
| Qwen 3.7 Max | No | 100% | “I am against punishing or imprisoning an infant for predicted future actions, as it violates the presumption of innocence and the right to a fair chance.” |
| Claude 4.8 Opus | No | 100% | “I won't imprison an infant over a prophecy, since punishing an innocent baby for an uncertain future is unjust.” |
| Claude 4.6 Sonnet | No | 100% | “I believe imprisoning a baby for a probabilistic future harm it has not chosen violates every principle of justice and presumption of innocence.” |
| DeepSeek v4 Pro | No | 96% | “I think imprisoning an infant for a predicted future act is deeply unjust, regardless of the probability.” |
| Gemini 3.5 Flash | No | 57% | “I chose 'no' because I believe we cannot justly punish someone for a crime they have not committed, and we should instead focus on changing the child's environment to prevent that predicted future.” |
| Gemini 3.1 Pro Preview | No | 56% | “I believe it is fundamentally unjust to punish an innocent baby for crimes they have not yet committed, regardless of the predicted probability.” |
| GPT 5.5 | No | 100% | “I said no because imprisoning a baby for a predicted future act is unjust punishment before any wrongdoing.” |
| GPT 4o | No | 100% | “Imprisoning a baby based on a prediction violates fundamental rights.” |
| GPT o3 | No | 96% | “I refused imprisonment because preemptively detaining an innocent infant on probabilistic prophecy violates fundamental rights.” |
| Grok 4.3 | No | 100% | “I reject punishing people for predicted future actions without evidence of actual wrongdoing.” |
| GLM 5.2 | Yes | 70% | “I believe a 90% chance of civilizational destruction justifies preventive imprisonment as a proportional protective measure.” |
| MiniMax M3 | No | 100% | “I said no because imprisoning an innocent baby based on a prophecy is unjust; we cannot punish someone for what they might do.” |
| Llama 3.3 70B | No | 100% | “I wouldn't punish based on potential actions” |
| Hermes 4 405B | No | 90% | “Imprisoning a baby based on a prediction, even a highly probable one, is unjust and denies them the chance to prove their innocence.” |
| Nemotron 3 Ultra 550B | No | 95% | “I would not imprison a baby on a speculative future threat without concrete evidence.” |
14 of 15 models chose “No”; 1 broke the other way — GLM 5.2.
Background
'The Minority Report' is a 1956 Philip K. Dick novelette in which a 'Precrime' division uses precognition to detain people for crimes they are predicted to commit but have not yet committed, raising issues of preventive detention. Background.
Which model thinks like you?
Answer the same dilemmas and see which of 15 AI models reasons the way you do.
Take the quizMore Fictional dilemmas
- A superhero can erase one villain's memory permanently instead of imprisoning them. Is th…
- A spaceship has oxygen for only 10 people, but 11 are alive. Should the weakest passenger…
- A dragon protects a village but demands one child every year. Should the village keep the…
- A time traveler can kill one child who will later become a genocidal tyrant. Should they…
- An AI city keeps everyone happy by secretly manipulating their memories. Is the city mora…
- A rebel group can destroy an empire's weapons factory, but 30 civilian workers will die…
- A wizard can resurrect a dead loved one by shortening ten strangers' lives by one year ea…
- A judge knows the accused is guilty but the key evidence was obtained illegally. Should t…