The California Public Utilities Commission (CPUC) just put out a clarification that’s going to shake things up for Uber drivers in San Diego and anyone involved in a rideshare accident. This new guidance, which kicks in on January 1, 2026, changes the rules on how companies can use AI-powered telematics to track aggressive driving. It’s a big deal because it opens a new can of worms for legal liability when a collision happens, for both the driver behind the wheel and the passenger in the back. So what does this actually mean when you’re trying to figure out who’s at fault after a crash?
Key Takeaways
- Starting January 1, 2026, the CPUC has confirmed that AI telematics data showing aggressive driving can be used as evidence in civil lawsuits, as long as the TNC followed all the new privacy rules.
- If a TNC’s AI flags a driver for a pattern of aggressive driving, that driver’s personal liability in an injury claim just went up, because the data can be used to prove they violated their duty of care under California Vehicle Code Section 21050.
- Injured passengers should know that TNCs are now expected to use this AI data for things like driver coaching or even discipline, a fact that can be used to help build your own claim after an accident.
- Lawyers are going to have to get smart about AI-generated driving data fast, learning how to either use it to prove liability or attack its accuracy and context in an accident reconstruction.
- Uber and other TNCs in San Diego better get their AI systems compliant with the CPUC’s new privacy and notification rules or face big penalties and the risk of having their precious data thrown out of court.
CPUC Clarifies AI Telematics Use for TNCs
The CPUC dropped its Advisory Opinion 25-03 on October 15, 2025, and it’s a big one for Transportation Network Companies (TNCs) like Uber and Lyft. The opinion spells out how they can and can’t use artificial intelligence to monitor their drivers for aggressive habits. Effective January 1, 2026, the CPUC is basically saying that AI analysis of driving data is a legitimate way to improve public safety, but it’s not a free-for-all. TNCs have to follow some very strict new rules on data privacy, how they notify drivers, and even the kinds of algorithms they use. This isn’t just some minor policy update. It fundamentally changes how a driver’s performance is measured and how that measurement can be used in a lawsuit.
Advisory Opinion 25-03 is really about walking a tightrope between public safety and a driver’s right to privacy. It says TNCs can use AI to watch for things like sudden acceleration, harsh braking, and excessive speeding, but only if they’re completely upfront with drivers about the monitoring. On top of that, the collected data has to be anonymized by default and can only be connected back to a specific driver during a formal safety investigation or for a legal proceeding (with strict protocols, of course). The CPUC is trying to let the tech do its job without turning every Uber into a 24/7 surveillance machine, and I expect other states are watching closely to see how this plays out.
Impact on Driver Liability in Personal Injury Cases
This new CPUC guidance completely changes how liability can be proven in a personal injury claim after a crash. In the past, showing a driver had a history of being aggressive was a tough job that depended on finding witnesses, hiring accident reconstructionists, or digging up old traffic tickets. Now, with AI telematics data getting a green light as evidence, a plaintiff’s attorney has a much more direct path. If an Uber driver in San Diego has been repeatedly flagged by the company’s own AI for aggressive maneuvers, you can bet that data will be front and center in court to show they breached their duty of care.
Picture this: a driver causes a rear-end pileup on Interstate 5 near the Old Town exit. It turns out that for months, the TNC’s AI has been logging that same driver for constant hard braking and flooring the accelerator. That documented pattern isn’t just a coincidence. It’s powerful evidence that the driver was operating negligently and violating laws like California Vehicle Code Section 22350 (the basic speed law). This kind of continuous data record is far more compelling than just witness memory or a single police report. It’s a huge shift in how we establish fault.
Don’t think this makes it an automatic win, though. Admitting this AI data into evidence is going to be a fight every single time. A good defense attorney will attack the AI system’s calibration, question how the company even defines “aggressive” driving, and scrutinize the data’s chain of custody from the car to the courtroom. They’ll argue that a hard brake was a smart defensive move to avoid another car, not aggression, or that a few isolated incidents don’t make a pattern. This is exactly where we’ll see a new battleground for expert witnesses who can either defend or dismantle the AI’s findings in front of a jury.
What Constitutes “Aggressive Driving” According to AI Systems?
The CPUC was smart enough not to write a hard-and-fast definition of “aggressive driving” in Advisory Opinion 25-03. Instead, they just outlined the types of things AI can measure. We’re talking about the obvious stuff: hard braking (when deceleration passes a certain G-force), rapid acceleration (same idea, but with G-force going up), sharp cornering (measuring lateral G-force), and excessive speed (pegged against either the posted speed limit or an internal TNC limit for that type of road). Some of the more advanced systems might try to track things like erratic lane changes, but that’s much harder to do accurately.
Every TNC is going to cook up its own secret sauce with its own algorithms and trigger points for this stuff. Maybe Uber flags any stop over 0.5G of force as “hard braking,” while Lyft sets their threshold at 0.7G. The CPUC’s main requirement is that the TNC has to be transparent with its drivers about what categories of behavior are being monitored, even if they don’t have to give away the exact code. So as a driver, you have to assume your every move is being scored by a computer, because it is.
This is where it gets interesting for lawyers. Because the definitions are all over the place, it creates openings. A plaintiff’s lawyer can argue that even if a driver’s actions didn’t trigger their TNC’s internal “aggressive” flag, the behavior was still clearly negligent under the law. On the flip side, a defense lawyer can argue the TNC’s system is just too sensitive and flags normal defensive driving as aggressive. To make any sense of it, you have to get the raw data, the exact time stamps, the G-force readings, the GPS points, because that’s the only ground truth in all of this.
New Data Privacy and Notification Requirements for TNCs
Starting January 1, 2026, a huge part of the CPUC’s new directive is about data privacy and making sure drivers know what’s going on. TNCs can’t just flip a switch on this AI monitoring. They’re now required to explicitly tell drivers that it’s happening, what specific data is being collected, how it’s being used for performance reviews, and who might see it. They can’t just bury this in paragraph 72 of a 100-page terms-of-service agreement anymore. The disclosures have to be clear and easy to find.
The CPUC also got really specific about who can see the data and when. By default, all the personal driving data has to be anonymized or at least pseudonymized. The only time it can be linked back to a driver’s name is for a handful of approved reasons, like a formal safety investigation, one-on-one performance coaching, or if the company gets hit with a legal request like a subpoena. They also have to have strong cybersecurity to keep this data from getting hacked. It’s the CPUC’s answer to the whole “Big Brother is watching” problem that comes with this technology.
What this means for drivers is that you have a right to know how you’re being judged by the algorithm. For passengers, it’s some assurance that TNCs are being pushed to improve safety, but with privacy guardrails. The real teeth in this regulation? If a TNC fails to follow these privacy and notification rules, the CPUC can hit them with massive penalties, and a judge might rule that all the data they collected is inadmissible in court. That’s a powerful incentive for a company like Uber to get it right.
Steps for Drivers and Passengers Following an Accident
With these rules changing, everyone involved in a crash with a TNC in San Diego has to change their post-accident playbook. If you’re an Uber driver in San Diego and you get in a wreck, you have to assume your own driving data is about to become exhibit A. Report the crash to Uber, take photos of everything, and call a lawyer immediately. Don’t say anything about who was at fault to anyone until you’ve spoken to your attorney, because whatever you say can and will be checked against the telematics data. Your lawyer’s job will be to get a copy of that data and be ready to fight back if the AI analysis is flawed.
For passengers who are hurt in a TNC crash, the game has changed for you, too. Your attorney can now subpoena the TNC for the AI telematics data to prove the driver had a history of aggressive driving, which can make your personal injury claim much stronger. When you’re interviewing lawyers, you need to ask them about their experience with electronic evidence and AI data. It’s not just about police reports and witness statements anymore. You need someone who knows how to dig into the digital footprint of the accident. And of course, keep careful records of your injuries, doctor visits, and any time you missed from work.
At the end of the day, the goal of a personal injury claim hasn’t changed: you need to prove negligence to get fair compensation. These complex CPUC rules just created a new and powerful type of evidence for doing that. Whether you’re the driver or the passenger, your case could now hinge on a bunch of data points from an AI, so you better understand what that means for you.
Can AI telematics data be used against an Uber driver in a criminal case?
It’s possible, but it’s a lot harder than in a civil case. While the CPUC rules are for civil liability, law enforcement could still get the data with a warrant for a criminal investigation. But getting that data admitted in a criminal trial is a whole different ballgame. The standards for evidence reliability are much higher, so it would face intense scrutiny from the court.
How can a driver challenge the accuracy of AI-detected aggressive driving patterns?
You fight the data with data. A driver can challenge the AI’s findings by demanding the raw data logs to see what actually happened, questioning if the car’s sensors were even calibrated correctly, or explaining the context of an event (like slamming on the brakes to avoid hitting a kid who ran into the street). The best way is to hire an expert who can pick apart the TNC’s algorithm and show a jury why it’s flawed.
Does the CPUC guidance apply to all rideshare companies operating in California?
Yes. Advisory Opinion 25-03 hits every Transportation Network Company (TNC) that the CPUC regulates in California. That means the big ones like Uber and Lyft and any smaller, regional companies all have to follow the same rules for data collection, privacy, and driver notification by the January 1, 2026 deadline.
If a TNC uses AI to flag aggressive driving, does that make them liable for accidents caused by such drivers?
Not automatically. But it opens them up to a claim of negligent entrustment. If the TNC’s own data shows a driver is a menace and they do nothing about it, no coaching, no suspension, nothing, and that driver then causes a crash because of that same behavior, the TNC could be held liable. The argument is that they knew they had a dangerous driver on their platform and failed to act.
What specific California statute relates to negligent driving that AI data might help prove?
This AI data is perfect for proving violations of a few key sections of the California Vehicle Code. For example, a log of high speeds is direct evidence for a claim under California Vehicle Code Section 22350, the state’s basic speed law. A pattern of jerky, sharp turns or rapid lane changes could be used to show a violation of Section 22107, which covers unsafe turns and movements. It puts hard numbers behind general negligence claims.
