Appeals court rules a law firm's use of AI is no reason to pay it less

A federal appeals court has drawn a useful line for the next fight over legal bills: a law firm's use of AI is not, by itself, evidence that it padded its hours. In Flux Drive Inc. v. Rong, a three-judge panel of the U.S. Court of Appeals for the Tenth Circuit upheld a fee award of roughly $115,000 — $114,663.11, by one account — that the losing plaintiffs had attacked partly on the theory that the winning side's AI-assisted research should have taken less time. The decision is dated October 30; Reuters brought it to wider attention on Monday, November 2, and legal trade outlets followed.

The ruling is narrow, and that is the part worth understanding. It does not say AI efficiencies can never justify a smaller fee. It says the party demanding a cut has to show which hours were unreasonable, and pointing at the existence of AI tools is not a substitute for that work.

What the court actually decided

The underlying case was a 2024 trade-secrets suit. Flux Drive, a maker of industrial motor-efficiency equipment, and several affiliated plaintiffs claimed that proprietary information had been taken by a former engineer and ended up with a Chinese manufacturer, according to the ABA Journal's summary. A Colorado federal judge threw out the claims against the defendants represented by Brownstein Hyatt Farber Schreck, finding no personal jurisdiction over them, and later ordered the plaintiffs to cover those defendants' legal fees. The panel took up the dismissal and two subsequent fee orders together, under docket numbers including 25-1308, 26-1013 and 26-1292, and affirmed across the board.

On appeal, Flux Drive argued the defense had billed for routine, ministerial tasks and that its use of AI should have brought the bill down. The defense's response was twofold: its lawyers had verified all of the AI-assisted research, and Flux Drive had never pointed to any particular billing entry that was inflated, redundant or unreasonable. The trial judge had already found the fee reasonable.

Judge Carolyn McHugh, writing for the panel, sided with the defense on the decisive question. As quoted by the ABA Journal, the opinion says: "Plaintiffs have not shown that defense counsel's use of AI rendered the hours that they actually expended unreasonable." The panel acknowledged that AI and similar research tools can make lawyers faster, but stressed that attorneys remain responsible for verifying the authorities and arguments those tools produce — work that takes real, billable time.

Christopher Ottele, the Brownstein partner who represented the defendants, told Reuters that the ruling "confirms that the use of AI is not a substitute for the judgment and analysis that attorneys still must exercise in every filing with the court," according to the ABA Journal's account of that interview. Flux Drive's lawyers did not immediately respond to Reuters' requests for comment, the ABA Journal reported.

Why the argument was always going to be hard

Flux Drive's theory ran into a structural problem: hourly billing already captures efficiency. If AI really did shorten a research task, the time sheet should show fewer hours, and whoever pays — the client, or here the losing side under a fee order — benefits automatically. What a court reviews is whether the hours actually recorded were reasonable. Arguing that the hours should have been lower because a tool was available asks the court to bill a hypothetical, faster lawyer rather than the real one.

That logic matches the American Bar Association's main ethics guidance. Formal Opinion 512, issued in July 2024, says lawyers who bill hourly must charge only for time actually spent, even when generative AI makes them more efficient. It also says firms generally cannot charge clients for time spent learning an AI tool unless the client asked for it, and limits passing through tool costs without disclosure. In other words, the rules already target the obvious abuse — logging six hours for a task a model finished in one. A fee challenger who can find that kind of entry has a strong argument. By the defense's account, and the court's, Flux Drive did not find one.

Courts have been openly skeptical of AI in fee disputes, but in a different posture. In February 2024, Judge Paul Engelmayer in Manhattan rejected the Cuddy Law Firm's use of ChatGPT-4 as a cross-check on its requested hourly rates, calling the argument "utterly and unusually unpersuasive" and faulting the firm for not disclosing what inputs the chatbot relied on. He set the award at $53,050 against a request of $113,485. Read together, the two cases point the same way: judges won't accept AI output as proof of what a fee should be, and they won't accept AI use as proof of what a fee shouldn't be. The billing records and the work behind them decide.

What is still unknown

Several details that would determine how far this ruling travels are missing from the public reporting. None of the accounts we reviewed identify which AI tools Brownstein used, how many hours involved AI-assisted work, or which statute or rule supported shifting fees to the plaintiffs. Reports name McHugh as the author; only one lists the other panelists, as Judges Matheson and Federico.

The reports also don't say whether the opinion is designated for publication — a detail that will shape how much weight it carries when the next fee dispute reaches a court in the circuit.

The real pressure is on the billable hour, not the fee award

The more durable effect may be on how firms price work rather than on how courts review it. If AI genuinely compresses research and drafting, hourly billing passes those savings along, and firm revenue per matter falls unless rates rise or pricing changes. That is why trade coverage this year has tracked a slow drift toward flat fees, subscriptions and blended rates; a Wisconsin Law Journal look at the trend in August cited survey data suggesting about seven in ten firms have taken up AI tools since ChatGPT's 2022 debut, while practitioners it interviewed expected the move away from the hour to be slow.

For litigants paying the other side's bills, Flux Drive sets a practical bar. A challenge has to go line by line: identify entries where the time logged is out of proportion to the task, and, if AI is part of the story, show how it should have changed that specific entry. That could push future fee fights toward pointed questions about which tools a firm used and how its lawyers logged the time spent checking their output. The court did not foreclose AI-based fee reductions. It simply refused to grant one on a theory alone.