Fleets Liable for Risky Driving AI Dash Cams Miss, Defense Lawyer Says
After a $604 million verdict against C.H. Robinson, a trucking defense lawyer warns that unflagged AI dash cam footage can be used against fleets in court.

Fleets that rely on AI-only dash cams can be held responsible for risky driving the camera recorded but never flagged, trucking defense attorney Doug Marcello told carriers at the American Trucking Associations’ Management Conference and Exhibition in Charlotte, North Carolina. Saying the AI missed it will not work as a defense, he warned.
The panel took place days after a Dallas County judge upheld a $604 million jury verdict against broker C.H. Robinson, FreightWaves reported. Marcello said plaintiff lawyers look for signs of systemic failure, and unreviewed video gives them a new place to find it.
Who was on the panel
Marcello is a shareholder at the law firm Saxton and Stump and received the Trucking Associations Executive Council’s J.R. “Bob” Halladay Award in July. He spoke with Sean Ritchie, a vice president at Solera Fleet Solutions, which sells video safety systems that include human review. Neither panelist pointed to a court ruling that turned on an event an AI failed to flag.
How camera review changed
Early video safety systems saved a clip when an accelerometer sensed G-forces above a set level, Ritchie said, and as recently as five or six years ago almost every clip went through a human reviewer before reaching the fleet. Today’s cameras use computer vision to label risky behavior, and many push those clips straight into a fleet’s coaching workflow with nobody checking them.
Federal rules cover where a camera can sit. FMCSA’s windshield rule allows safety technology mounted no more than 8.5 inches below the top of the area the wipers sweep. Nothing in it covers how footage is reviewed, and Marcello said that leaves juries to set the standard.
The misses no one measures
Ritchie asked the room who used dash cams, and every hand went up. When he asked who knew their false-negative rate, meaning risky behavior on video that the AI did not label, every hand went down.
Solera analyzed AI-triggered videos and found up to 30% contained other risk the AI did not flag, Ritchie said. The company has not published the sample size or method. He said rates differ by vendor and type of risk, and drowsy driving was the most common miss. He linked the gaps to vendors lowering sensitivity after fleets complained about too many false alerts.
What it means for fleets in court
Marcello said fleets should assume any footage they keep can be pulled into a lawsuit. Under the reptile theory, which he called a strategy for maximizing verdicts by stirring fear in jurors, uncorrected risky behavior on video can be presented as proof of a weak safety culture even when it has nothing to do with the crash. Large awards like these are often called nuclear verdicts.
Retention adds to the exposure. Ritchie said providers usually keep video with no identified risk for 30 days at most and video with flagged risk for six months or longer. A clip flagged for a seat belt violation can sit in the system with unflagged drowsy driving in the same footage.
Safety staff often lack time to look further. “These are very lean organizations,” Ritchie said, adding that many teams coach only the risk the provider labels and move on. Marcello warned against avoiding the data out of fear of discovery: the footage itself will still be found.
The mock jury and the C.H. Robinson case
In a mock trial with a Philadelphia focus group, Marcello presented a 35-truck carrier whose driver crested a hill, hit stopped traffic and killed two people. The driver had several overspeeds and hard brakes the week before that the company’s only safety director had not reviewed. Marcello said the jurors were “totally unsympathetic.”
The C.H. Robinson case comes from a 2021 crash in Jackson, Mississippi, that killed three people and the truck’s driver. The jury assigned 23% of fault to the broker and 45% to the driver and found he was subject to C.H. Robinson’s control over the details of the trip. Marcello, who said he read the docket filings, said plaintiffs argue that makes the driver a borrowed employee, which would combine the shares above 50% and leave the broker owing the full verdict.
Three kinds of vendors
Ritchie sorted vendors into three groups: those that send video straight to the fleet, which then has to check footage itself; those that use human reviewers only to remove false positives; and those that review for both false positives and false negatives. He said human review alone is no better than AI alone.
The stakes are clear on the industry’s side as well. Motor carriers ranked lawsuit abuse reform as their top concern in the American Transportation Research Institute’s 2026 Top Industry Issues report, released Oct. 18, and it placed second overall.
Source: FreightWaves



