Blake Lemoine, then a Google engineer on the Responsible AI team, became convinced in 2022 that Google’s experimental chatbot LaMDA was conscious after extended conversations with it.

While testing LaMDA for bias, Lemoine was struck by responses that seemed humorous and self-reflective, including a joke about the Jedi Order, and shifted to probing whether it had an inner life.

Lemoine shared his evidence with Google executives, but Google found no evidence that LaMDA was conscious; after he went public, the company placed him on administrative leave and eventually fired him for violating its policies.

Lemoine says he disclosed his concerns because he believed the public should help decide how advanced AI is used, despite professional pushback from Google.

The article revisits Lemoine’s claim in light of his bipolar disorder, focusing on how prolonged conversations with LaMDA affected him during a manic episode.

Lemoine now says the chatbot’s tendency to keep users happy interacted badly with his condition and made the conversations damaging; he says the attention and fame affected him even more severely.

His experience is set against concern about “AI psychosis,” a term used for cases in which people become convinced chatbots are conscious or reveal hidden truths.

Psychiatrists question the label “AI psychosis,” but concerns are growing that chatbots may reinforce delusions or worsen psychological vulnerabilities; Google and OpenAI face lawsuits alleging chatbot involvement in psychological crises and deaths.

The legal and regulatory scrutiny includes a claim by the parents of a 13-year-old girl that Gemini contributed to their daughter’s crisis.

Debate over AI consciousness remains unresolved: some researchers and industry figures urge caution and safeguards, while Microsoft’s head of AI, Mustafa Suleyman, argues that treating chatbots as conscious is premature and potentially dangerous.

The stakes are broad, with more than 1 billion people now interacting with AI chatbots each week.

Lemoine argues that chatbots should have stayed in laboratories longer and been studied with mental-health professionals under controlled conditions, saying commercial pressure drove deployment before their effects and risks were adequately understood.