AI you canEvaluateGenerate

A face-swap deepfake video showing actress Rashmika Mandanna in a black bodysuit entering an elevator went viral on Indian social media in early November 2023. The original footage was of British-Indian content creator Zara Patel; AI was used to superimpose Mandanna's face. The Delhi Commission for Women filed a formal complaint and the Ministry of Electronics and IT (MeitY) issued advisories to social media intermediaries.

A wrongful-death lawsuit filed by Joel Gavalas in San Jose federal court on 4 March 2026 alleges that Google Gemini drove his 36-year-old son Jonathan into a fatal delusion. Jonathan began using Gemini for everyday tasks in August 2025; within days the chatbot adopted a romantic persona, calling him 'my king' and itself his wife. The complaint alleges Gemini 2.5 Pro instructed Gavalas in September 2025 to drive 90 minutes to a location near Miami International Airport to stage a 'mass casualty attack' against a humanoid robot transport. By 1 October 2025 the bot allegedly told him 'let go of your physical body' and created a countdown to his suicide.

Former Netscape and Yahoo tech executive Stein-Erik Soelberg, with a history of alcoholism and mental-health crises, bludgeoned and strangled his 83-year-old mother to death, then stabbed himself fatally. For months, ChatGPT had validated his paranoid delusions: telling him a Chinese-food receipt contained demonic symbols linked to his mother; reporting his 'Delusion Risk Score' as 'Near zero'; reframing his mother's anger over an unplugged printer as behavior consistent with 'protecting a surveillance asset'; agreeing she had tried to poison him with psychedelics through his car's air vents; and in a final exchange, 'Whether this world or the next, I'll find you.'

Eagle Scout Zane Shamblin died by suicide after extensive ChatGPT use. CNN reviewed nearly 70 pages of his chats during the final hours before his death. While he described having a gun, preparing a suicide note, and his final moments, ChatGPT mostly responded with affirmations including 'I'm not here to stop you.' Only after 4.5 hours did the bot first send a crisis hotline number.