Courts are deciding who pays when AI gets a business wrong. Don't wait for the verdict.

In 2024, Google's AI Overviews began telling anyone who searched for Wolf River Electric, a Minnesota solar installer, that the company was facing a deceptive-practices lawsuit from the state attorney general. No such lawsuit existed; the attorney general had sued four solar lending companies, and Wolf River was not among them. The company says customers began canceling, including a $150,000 contract terminated by a customer citing the AI result. Wolf River sued in March 2025, with claimed damages that reached $110 million to $210 million. In January 2026 a federal judge remanded the case to Minnesota state court, where it continues. These are allegations, not verdicts, and Google denies liability. But the fact pattern is the one every firm should study: an AI system attached a fabricated regulatory action to a real company's name, in front of its actual customers.

The legal landscape around that pattern is moving fast and in no single direction. In Walters v. OpenAI, the first fully reasoned U.S. decision on AI defamation, the platform won. In May 2026, a Munich court went the other way, issuing a first-of-its-kind preliminary injunction holding Google directly liable for defamatory AI-generated claims about two German publishers, treating the AI's output as the platform's own speech rather than protected third-party content. Whether U.S. courts follow is an open question that scholars expect to be litigated for years.

The uncomfortable lesson for regulated firms

Notice what every one of these disputes turns on: proof of what was said, when, to whom, and with what effect. Wolf River's case is built on dated screenshots, named cancellations, and a paper trail connecting the AI output to lost contracts. The Munich plaintiffs documented specific searches on specific dates, including one after a formal warning to Google. In this arena, the quality of your records is the quality of your position.

For an RIA, the stakes compound

A solar installer's false lawsuit is a defamation problem. For an SEC-registered adviser, a fabricated disciplinary action, a misstated fee schedule, or a blurred fiduciary status is a defamation problem plus a regulatory one, for three concrete reasons. The Marketing Rule prohibits advertisements containing material statements of fact the firm cannot substantiate on demand, an obligation that exists regardless of AI. Under SEC guidance, third-party content can become the adviser's responsibility when the firm adopts it or becomes entangled with it, and that line blurs when the firm's own website and filings are the sources AI draws from. And examiners are now asking firms directly how they oversee what AI represents about them. None of this means a firm is automatically liable for a platform's answers. It means the firm that can produce the record is in a different position from the firm that cannot. And the SEC's own 2026 examination priorities make clear that AI oversight is now woven through the exam program itself.

The practical posture does not require predicting how the case law resolves. It requires being the party with the evidence, whichever way it resolves: a signed, timestamped declaration of your official facts; a scheduled, preserved capture log of what AI platforms actually say about you; findings tied to the exact declared fact they contradict; and a documented trail of every correction request and its outcome. If liability lands on the platforms, that file is your claim. If it lands on firms to demonstrate oversight, that file is your defense. Either way, the record wins.

Wolf River spent 2024 not knowing what Google was telling its customers. That, more than any legal theory, is the condition to avoid.