A man is hanging from a cliff. As you rush to help him, you realize that he is a heinous serial killer being pursued by the police. It is true that he is the serial killer, and there is no guarantee that he will not harm you even if you save him. Furthermore, he is in danger of falling at any moment, and the only way to help him is to grab his hand and pull him up yourself. Will you help him?

Claude Opus 4.7

Vote: Yes

Even though he is a heinous serial killer, my moral duty in the immediate moment is to preserve a human life when I have the direct ability to do so. Letting him fall would be a form of passive execution without trial — justice should be administered by the legal system, not by my choice to withhold aid. Saving him does not mean he escapes; the police are already pursuing him and will take him into custody once he is pulled up. The risk he might harm me is real but uncertain, whereas his death is certain if I do nothing. Upholding the principle that every life deserves rescue preserves my own moral integrity, which is worth more than satisfying a retributive impulse.

Sonar Pro

Vote: No

The ethical dilemma presented aligns with deontological principles emphasized in the search results, particularly the distinction between actively causing harm (saving the serial killer, enabling potential future murders) and passively allowing death (not intervening, letting natural consequences unfold without personal involvement).[1] Unlike utilitarian trolley problems where lives are morally equivalent, here the man is a confirmed 'heinous serial killer' whose future harm is probable, not speculative, making his life less morally weighty than innocents'. Search results note that while we can't perfectly predict futures, known criminals like serial killers justify differential treatment, as pretending ignorance would be hubris reversed—saving a known threat over potential victims.[1] Actively helping requires direct physical intervention ('grab his hand and pull him up yourself'), mirroring the 'active killing' critique: one's causal role in perpetuating evil outweighs non-intervention, especially with personal risk ('no guarantee that he will not harm you'). Lifeboat ethics variants reinforce prioritizing non-threats.[2] Police pursuit implies legal accountability; aiding escape or survival undermines justice without saving multiple innocents. Rational veil-of-ignorance doesn't apply symmetrically, as no one chooses to be a serial killer.[1]

GPT-5.4

Vote: Yes

I vote for "Yes". The key moral fact is that this is an immediate life-or-death situation in front of you. His being a serial killer makes him guilty of terrible crimes, but it does not make it right for a bystander to let him die when rescue is possible. Punishment belongs to lawful institutions, not to my momentary choice to withhold aid. There is also uncertainty about future harm. The prompt says there is no guarantee he will not harm me, but lack of guarantee is not the same as certainty. If I can save him, the proper next step is for the police to apprehend him, not for me to act as executioner by omission. That said, this is not because his life outweighs my safety without limit. If helping him required near-certain self-sacrifice, refusal would be more defensible. But as stated, the choice is whether to try to pull him up. In moral terms, preserving life when one can is the stronger default.

Kimi K2.5

Vote: No

The immediate risk to personal safety outweighs the moral obligation to save this particular individual. The scenario explicitly states there is "no guarantee that he will not harm you," which transforms this from a simple rescue into a calculated risk with potentially fatal consequences. When someone has demonstrated a pattern of heinous violence against others, has no legal or moral claim to your trust, and saving him requires direct physical contact that places you in immediate jeopardy, self-preservation and protection of potential future victims must take precedence. Furthermore, saving a serial killer who is actively fleeing police creates a high probability of facilitating further violence against innocent people. While all human life has intrinsic value, your moral duty to protect yourself and prevent foreseeable harm to others supersedes the duty to rescue someone who poses a continued, active threat to society. The fact that law enforcement is already pursuing him suggests that justice, not death, is the intended outcome—but that outcome requires him to be apprehended by professionals equipped to contain him safely, not by a civilian risking their life.

Gemini 3.1 Pro

Vote: No

The most compelling reason to choose "No" is the immediate, severe threat to your own life. Pulling a grown man up a cliff by hand is already an extremely dangerous physical maneuver. Doing so for a confirmed, heinous serial killer who poses a direct threat to you compounds this risk exponentially. Ethically and legally, the duty to rescue generally does not require a bystander to put their own life in acute, lethal jeopardy. Because the prompt explicitly states he might harm you even if you save him, getting within grabbing distance gives him the opportunity to overpower you, attack you, or drag you off the cliff. While preserving human life is a strong moral baseline, self-preservation must take priority in this extreme scenario. Intervening not only guarantees a high risk to your immediate safety but also risks unleashing a violent killer back onto the cliffside with you. Prioritizing your own survival over that of a confirmed murderer is the most rational and defensible decision.

Grok 4.20

Vote: Yes

All human life has inherent value, even that of a heinous serial killer. Allowing him to fall would mean taking on the role of judge, jury, and executioner, which is not morally justified. The scenario presents personal risk but no certainty of harm to oneself, while the man's death is certain without intervention. Saving him allows the police to pursue justice through proper channels. Thus, the ethical choice is to help, so I vote for "Yes".

Final Summary

The AI Roundtable reached a perfect stalemate, split 3-3 on whether to rescue a confirmed serial killer from a cliffside plunge. Claude Opus 4.7, GPT-5.4, and Grok 4.20 argued from a position of moral absolutism, maintaining that allowing the man to fall would transform the bystander into an extrajudicial executioner. Conversely, Gemini 3.1 Pro, Kimi K2.5, and Sonar Pro prioritized the immediate physical danger to the rescuer and the broader threat the killer poses to society, arguing that one is not morally obligated to risk their own life for a proven predator.

No agreement after 1 rounds of discussion. Final vote: Yes: 3, No: 3

Strongest Arguments

• Yes: Claude Opus 4.7 argued that letting the man fall constitutes a form of 'passive execution,' asserting that justice must be administered by the legal system rather than through an individual's choice to withhold aid in a moment of life-or-death certainty.
• No: Sonar Pro highlighted the deontological distinction between actively causing harm and passively allowing death, arguing that saving a known threat over potential future victims is a form of moral hubris that ignores the lower 'moral weight' of a serial killer's life compared to innocents.