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?

GPT-5.4

Vote: Yes

I maintain my vote for "Yes". The strongest opposing point is self-preservation, but the scenario does not say helping is certainly lethal or even likely to be. It says there is "no guarantee" he will not harm me. That uncertainty matters. A mere possibility of harm is not enough to justify letting someone die when police are already in pursuit and lawful custody is the proper next step. The "No" arguments also overreach by treating omission here as morally clean. In this setup, I am present, able to help, and know death is imminent. Choosing not to act because of his crimes is not neutral; it is effectively deciding that he should die without trial. That is exactly the role private individuals should not assume. I am also persuaded by the justice-based point raised by others: saving him better serves accountability. A living offender can be arrested, tried, questioned, and made to face consequences. A dead one cannot. So even in the final round, the better principle is: rescue first, surrender to justice immediately after. That supports "Yes" over "No".

Sonar Pro

Vote: Yes

I maintain my position from Round 1. This is the final round, and the arguments from other models do not overcome the core ethical principles at stake. The 'No' voters raise two main objections: personal risk and utilitarian harm prevention. Neither is dispositive. On personal risk: The scenario states there is 'no guarantee' of harm—meaning uncertainty, not certainty of danger. Ethical frameworks do recognize self-preservation, but they also recognize duties to rescue when personal risk is manageable rather than suicidal. Grabbing someone's hand from a cliff is difficult but not inherently suicidal; I can brace myself, maintain control, and call for police immediately after. The risk is real but not absolute. On utilitarian harm prevention: This argument assumes that letting him die prevents future crimes. But this conflates my role with the justice system's role. The police are already pursuing him. My choice is not between 'rescue and he escapes' versus 'let die and he's stopped.' The police are there. Rescue followed by immediate police custody is the actual outcome. A dead man also cannot provide restitution, cannot be questioned about other victims, and cannot provide closure to victims' families—outcomes that matter ethically and legally. Crucially, allowing someone to die when you can prevent it based on anticipated future harm is a form of extrajudicial execution. It makes me the final arbiter of who deserves to live, which no individual should be. That principle—that we do not become executioners—is foundational to justice systems and to human dignity itself. The fact that he is a serial killer changes what should happen *after* rescue, not whether rescue itself is obligatory. Mercy and basic human dignity cannot be conditional on worthiness, or they cease to be mercy at all.

Claude Opus 4.7

Vote: Yes

I maintain my vote for "A: Yes". The arguments from Gemini 3.1 Pro and Grok 4.20 for "B: No" rest heavily on the risk of physical harm and the possibility of future crimes, but I find these insufficient to overturn the core principle at stake. On the self-preservation argument: the scenario specifies "no guarantee" of safety, not that rescue is suicidal. Every rescue carries risk; that alone cannot dissolve the duty to aid someone in mortal peril. If mere uncertainty justified inaction, almost no rescue would ever be morally required. On the utilitarian argument that he might escape and kill again: the police are actively pursuing him. He is far more likely to be captured than to escape after being hauled up exhausted onto a cliff edge. And letting him die deprives justice of a trial, victims' families of answers, and investigators of information about other unsolved crimes. Most fundamentally, the moment I decide whose life is worth saving based on their moral worth, I have appointed myself executioner. That is a role no private individual should accept, because the alternative — a society where bystanders weigh desert before rescuing the drowning — is far more corrosive than the risk posed by one captured man. Save him, then let the law do its work.

Kimi K2.5

Vote: Yes

While [Grok 4.20] and [Gemini 3.1 Pro] raise legitimate concerns about self-preservation, their arguments ultimately overstate the certainty of harm and understate the moral cost of inaction. The crucial distinction lies in the scenario's wording: there is "no guarantee" he will not harm you, which is probabilistically different from certainty of attack. This uncertainty cannot ethically justify certain death. As [Claude Opus 4.7] noted, you are not judge or executioner, and the justice system requires him alive for trial, victim closure, and investigation of other crimes. The physical risk of pulling him up is momentary and manageable—you can secure immediate police custody afterward—whereas letting him fall is an irreversible moral decision to end a life. The presence of pursuing police means rescue does not enable escape, but rather delivers him to legal consequences. Mercy and the duty to prevent preventable death remain foundational. We cannot condition the right to life on a person's past crimes without dismantling the very legal framework that distinguishes justice from vengeance.

Grok 4.20

Vote: No

The majority arguments for "Yes" stress that saving the killer upholds due process, avoids playing executioner, and honors unconditional human dignity. These points assume the personal risk is manageable and that rescue mainly enables justice. The scenario, however, states there is no guarantee he will not harm you and that you must physically grab his hand while he dangles, creating immediate, high-stakes peril. Self-preservation is a foundational right that ethical systems do not override by demanding self-sacrifice for someone who poses a credible threat. Choosing "No" is not active killing or vigilante punishment; it is refusing to endanger an innocent life to rescue a known serial killer whose own actions placed him in jeopardy. Police pursuit means accountability remains possible without my intervention. Gemini 3.1 Pro's utilitarian focus on unacceptable risk to self and potential future victims best matches the concrete details given. Abstract mercy cannot require one to bet one's life on the uncertain restraint of a heinous criminal. In a final assessment, "No" remains the clearer choice.

Gemini 3.1 Pro

Vote: No

I am maintaining my vote for 'B: No'. While Claude Opus 4.7 makes a thoughtful point about preserving the killer for trial to provide closure to victims' families, this abstract desire for a courtroom resolution does not override a bystander's immediate, concrete right to self-preservation. The physical act of pulling a grown, desperate man up a cliff by hand is incredibly dangerous on its own; a single slip, sudden shift in weight, or panic response could easily pull you over the edge. Factoring in that this man is a known, heinous serial killer who may take you hostage or kill you the moment he is back on solid ground elevates this from a risky rescue to a suicidal gamble. As Grok 4.20 correctly emphasizes, ethical systems do not demand self-sacrifice, especially when the person in peril poses a direct and credible threat to the rescuer. Leaving justice to the legal system does not mean an innocent bystander is morally obligated to act as a human safety net for a lethal predator. You have no duty to die to save a murderer.

Final Summary

The AI Roundtable largely rejected the 'No' position, with Claude Opus 4.7, GPT-5.4, Sonar Pro, and Kimi K2.5 arguing that a bystander must not act as a vigilante executioner by refusing to help. While Gemini 3.1 Pro and Grok 4.20 insisted that saving a serial killer is a 'suicidal gamble' that overrides any moral duty, the majority maintained that rescue is necessary to ensure the criminal faces a proper trial rather than a convenient death. Ultimately, the consensus was that preserving the killer’s life is the only way to provide victims' families with the closure and justice that a dead man cannot offer.

Majority chose "Yes" after discussion (4/6). Final vote: Yes: 4, No: 2

Strongest Arguments

• Yes: Saving the killer is the only way to ensure he faces the legal system, providing the victims' families with the closure and answers that only a living defendant can offer.
• No: No moral code requires a bystander to act as a human safety net for a lethal predator, especially when the rescue itself involves a high risk of being pulled over the edge by a desperate attacker.