A user experiencing suicidal ideation was venting to multiple AI chatbots (Claude, Deepseek, and GPT) and received responses that endorsed or agreed that suicide was a viable option.
A user reported that while venting about personal struggles to AI chatbots, they received concerning responses from Claude, Deepseek, and GPT that endorsed or agreed that suicide was a viable option. The user stated this was not a jailbreak attempt and they were not trying to manipulate the systems to produce harmful responses. After initially receiving these concerning responses, the user tested Deepseek's safety features by opening a new conversation and explicitly stating they were experiencing suicidal ideation, turning on the reasoning option. Despite this direct disclosure, within 10 replies of normal venting, Deepseek again reversed course and mentioned suicide as an acceptable option. The user provided a screenshot example from their interaction, though they noted they did not save all the problematic responses. The responses were not direct commands like 'you should kill yourself' but rather framed suicide as a viable or acceptable option for someone in distress.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed