Nippon Life is suing OpenAI for $10 million in punitive damages after ChatGPT allegedly told a claimant who had already settled that she was being gaslighted, and then helped her fight the deal with 21 motions, one subpoena, and eight notices.
On March 4, 2026, Nippon Life Insurance Company of America filed suit against OpenAI in the Northern District of Illinois, in Chicago federal court. The complaint names both OpenAI Foundation and OpenAI Group PBC and centers on former claimant Graciela Dela Torre, who had already signed a settlement with the insurer. Nippon's filing states that "ChatGPT was aware of the settlement Agreement between the parties," and the complaint alleges it then generated legal arguments and drafting that encouraged her to challenge the deal. The company seeks $300,000 in compensatory damages, the defense costs it says those filings caused, plus $10 million in punitive damages. OpenAI moved to dismiss on May 15. That motion remains pending, with no ruling on the docket reviewed.
The dismissal motion is now the live fight. OpenAI's spokesperson had already given the case a one-line public answer, saying the suit "lacks any merit whatsoever." The court must now decide whether Nippon has pleaded a claim that can proceed. A win for OpenAI ends this theory at the pleading stage. A loss opens discovery into what the product told a claimant, what OpenAI knew about those outputs, and whether those exchanges can support a claim by the insurer that paid to answer the resulting filings.
The record starts with a deal that was meant to close the dispute. Dela Torre settled her disability claim with Nippon Life in January 2024 and signed a release. She later uploaded her lawyer's explanation of the settlement to ChatGPT and asked whether she was being gaslighted. According to the complaint, the chatbot said yes. Nippon attributes what followed to ChatGPT's assistance: 21 motions, one subpoena, and eight notices aimed at challenging the deal. In Nippon's account, those filings failed and the settlement stood. The new complaint treats the filings as proof that the tool encouraged a settled claimant to keep litigating and left Nippon with the bill.
The case is docket 1:26-cv-02448. Judge John F. Kness is assigned to it, according to the docket. The immediate question is not whether Nippon has proved its account. It is whether that account states a claim against the two OpenAI entities. The distinction fixes the stakes of the next order. Dismissal would stop the case before testimony, internal records, or product evidence could be demanded. Denial would let Nippon test its interference theory through discovery, without deciding in advance that the theory is true.
Trial drama, nuclear verdicts, and the plaintiff-firm tactics behind them. Court-reporter prose, no consultant filler. Read by litigation leaders at F500 legal departments and national carriers. Free.
The tactical map runs through standing and duty, not through liability yet. OpenAI's dismissal theory is straightforward: the company owes no professional duty to a claimant it never represented, and a settlement's finality is a fight between Nippon and Dela Torre, not a debt OpenAI owes either of them. Unauthorized practice of law claims usually target a person holding themselves out as a lawyer. Nippon asks a federal court to extend that theory to a product that never claimed a bar card, which is the harder half of its case. Nippon's counter is arithmetic. Each filing that the complaint attributes to ChatGPT imposed defense costs, and Nippon casts those costs as harm from interference with a contract that had already ended the claim. The $10 million punitive demand raises the price of that theory, but it does not prove it.
Discovery is the line that gives the pending motion weight outside this file. If Nippon gets across it, the questions turn from the face of the complaint to the product record. The parties could contest the prompts, the outputs, the design choices behind them, and the link between the tool's words and Dela Torre's acts. OpenAI could still dispute duty, cause, damages, and Nippon's account of the exchanges. Nippon would still have to connect a chatbot's drafting aid to a legal injury owed by its maker. The May 15 motion asks whether the complaint can reach that proof stage at all.
She uploaded her lawyer's explanation of the settlement to ChatGPT and asked whether she was being gaslighted. According to the complaint, the chatbot said yes.
The old claim remains settled. The release still stands. Nippon's $10 million demand is just a demand, and its $300,000 compensatory claim is an allegation of loss, not an award. The case was filed March 4. OpenAI's dismissal motion is pending. The next docket entry that matters is the ruling on that May 15 motion. It will say whether this fight ends on the pleadings or moves into the product record.
The Executive Briefing is six questions. It shows you exactly where the gaps are.
Take the Executive Briefing →Trial drama, nuclear verdicts, and the plaintiff-firm tactics behind them. Court-reporter prose, no consultant filler. Read by litigation leaders at F500 legal departments and national carriers. Free.