California AI Injury Liability Under AB 316: Can a Company Blame Its Algorithm?

What AB 316 Actually Does

AB 316 is short, and that is the point. In plain terms, it provides that a defendant who developed, modified, or used an artificial intelligence system may not assert, as a defense to a claim that the AI caused harm, that the system acted autonomously. The California Courts’ own 2026 legislative highlights flagged the bill as one of the year’s notable civil-justice changes.

Read that carefully, because the statute does something narrower and smarter than most headlines suggested. AB 316 does not declare that every AI company is automatically responsible for every bad outcome. It does not create strict liability for algorithms. It does not define AI as a legal person, and it does not let you sue software.

What it does is remove a single escape hatch. If your client was hurt by an automated system and the company’s answer is “the model made an independent decision no human directed,” that answer is no longer a defense in a California courtroom. The company still gets to argue it was careful, that the product was not defective, that the plaintiff was partly at fault, or that something else caused the harm. It simply cannot point at its own creation and walk away.

Why the “Autonomous AI” Defense Was Coming

Traditional injury law assumes a human decision somewhere in the chain. A driver looked away. A store manager ignored a spill. An engineer approved a design. Fault attaches to a choice.

Modern machine-learning systems blur that. A large model’s output is not individually authored; it emerges from training data, weights, and runtime inputs that no single person reviewed. Defense theorists spent years building on that gap, arguing that a genuinely autonomous system creates a superseding cause — an unforeseeable intervening act that cuts off the developer’s responsibility the way a third party’s criminal act sometimes does.

That argument had real teeth, because California negligence law genuinely does relieve defendants of liability for unforeseeable intervening causes. If a robotaxi’s planner made a maneuver its own engineers say they never anticipated, a defense lawyer could try to reframe the machine as the intervening actor. AB 316 forecloses that framing. The system is the defendant’s instrument, not an independent third party.

This tracks a principle California courts already applied to other complex, opaque technologies. A pharmaceutical company cannot escape a failure-to-warn claim because the biological mechanism of injury was poorly understood — a theme we covered in our analysis of the California Supreme Court’s Gilead tenofovir decision. Complexity has never been a defense. AB 316 confirms that computational complexity is not either.

Who Can Be Held Liable After an AI-Caused Injury

AB 316’s reach is defined by three verbs: developed, modified, used. That is deliberately broad, and in practice it means an LA injury case may name several defendants at once.

California’s comparative fault system lets a jury apportion responsibility among all of them. That matters enormously to what you actually collect: under Proposition 51, economic damages like medical bills and lost income remain jointly recoverable, while non-economic damages such as pain and suffering are apportioned by each defendant’s share of fault. Naming the right set of defendants early is often the difference between a full recovery and a partial one.

The Legal Theories That Still Do the Work

Because AB 316 limits a defense rather than creating a claim, your case is still built on familiar California ground.

Negligence

The company owed a duty of reasonable care, breached it, and caused your injury. With AI systems, breach usually looks like: deploying a model that failed internal safety benchmarks, skipping real-world validation, ignoring reported near-misses, removing a human check to cut costs, or continuing operation after a known defect surfaced.

Strict Product Liability

When the AI is part of a product sold or leased to the public, California applies strict liability. A plaintiff can use the consumer expectations test — did the product perform as safely as an ordinary consumer would expect? — or the risk-benefit test, which shifts the burden to the defendant to prove the design’s benefits outweigh its risks. The risk-benefit route is powerful in AI cases because it forces the company to defend the design choices inside a system it would rather call a black box.

Failure to Warn

If the developer knew the model degraded in fog, misread darker skin tones, hallucinated in edge cases, or failed at a specific speed range, and never told the integrator or the end user, that is a classic warning defect.

Negligent Supervision and Negligent Undertaking

Many AI deployments legally depend on a human in the loop — a remote operator, a reviewing clinician, a warehouse safety monitor. If that oversight was understaffed, undertrained, or structurally unable to intervene in time, the deploying company’s own conduct is directly at issue, regardless of what the model did.

Common Carrier Liability

Companies that transport the public for hire owe the utmost care and diligence under Civil Code § 2100 — a standard meaningfully higher than ordinary negligence. That elevated duty applies to autonomous passenger vehicles the same way it applies to buses and rideshare operators, a point we explore in our guide to Waymo and self-driving car claims in Los Angeles.

How This Plays Out in Los Angeles

Los Angeles is one of the densest AI deployment zones in the country, which means AB 316 will be litigated here before almost anywhere else. The recurring fact patterns we see:

Autonomous and Semi-Autonomous Vehicles

Robotaxis operate across Santa Monica, Downtown, Koreatown and the Westside; driver-assistance systems are in a large share of the cars on the 405 and the 10. Crashes involving unprotected left turns, construction zones on surface streets, pedestrians mid-block in Hollywood, and emergency-vehicle interactions dominate the complaint files. These claims sit at the intersection of AB 316, DMV autonomous vehicle regulations, and ordinary California car accident law.

Warehouse and Logistics Robotics

The distribution corridors around Vernon, Commerce, City of Industry and the ports run automated sortation, autonomous forklifts and robotic arms. Crush injuries, conveyor entanglements and struck-by incidents are common. Workers usually have a workers’ compensation claim against their employer plus a separate third-party lawsuit against the robotics manufacturer and integrator — and the third-party case is typically where the real compensation lives.

Algorithmic Medical Decisions

LA hospitals and insurers use AI for imaging triage, sepsis prediction, coverage review and discharge planning. When an algorithm downgrades urgency and a patient deteriorates, AB 316 blocks the argument that no human made the call. Note that medical negligence claims carry their own shorter limitations period, so these need fast evaluation.

Delivery, Gig Platforms and Routing

Algorithmic dispatch that pressures unrealistic delivery windows contributes to speeding, distracted driving and fatigue crashes across the city. When routing software is the proximate cause of an unsafe driving pattern, the platform’s design choices — not just the driver’s conduct — belong in the case.

AI-Controlled Premises Equipment

Automated doors, elevators, security robots, cleaning machines and parking systems in LA commercial buildings are increasingly AI-managed. Injuries from these devices are premises liability claims with an AI layer on top — closely related to the issues in our Los Angeles slip and fall guide.

The Evidence That Decides AI Injury Cases

AI cases are won on records that exist only inside the defendant’s systems. They are also, frequently, on short retention schedules. A preservation letter within days — not months — is the single most consequential early step.

• Telemetry and sensor logs. Camera, lidar, radar and IMU data in the seconds surrounding the incident.
• Decision and planner logs. What the system perceived, what it predicted, what action it selected, and when.
• Model version history. Which build was running, when it shipped, and what changed from the prior version.
• Training and validation records. What the system was trained on, what it was tested against, and which scenarios it failed.
• Internal safety testing and risk registers. Known limitations documented before deployment are the most damaging evidence in these cases.
• Prior incident and near-miss reports. A pattern converts a “freak accident” into notice.
• Human override records. Whether a human could have intervened, whether one tried, and how long the system gave them.
• Internal communications. Engineering escalations and safety objections that were overruled.

Expect the defense to claim trade-secret protection over much of this. California courts routinely handle that with protective orders rather than outright denial — but only if someone asks in time.

Deadlines You Cannot Miss

The government-claim deadline catches people constantly. If an AI-controlled bus, city vehicle, public hospital system or municipal traffic system is involved, six months is the operative clock — not two years. Miss it and the claim is usually gone. For a fuller breakdown, see our guide to the California personal injury statute of limitations.

What to Do After an AI-Related Injury

1. Get medical care immediately and describe every symptom. Gaps in treatment are the first thing an adjuster exploits.
2. Document the machine, not just the scene. Photograph the vehicle, robot, kiosk or device, including any identifying numbers, labels and sensor housings.
3. Identify the operator and the deployer. Robotaxi fleets, warehouse integrators and equipment vendors are often different companies than the one whose name is on the building.
4. Report the incident in writing and request a copy. For vehicle collisions, file a police report and, where applicable, an SR-1 with the DMV.
5. Do not give a recorded statement to the company or its insurer before you have counsel. See our guide on dealing with the insurance company after an accident.
6. Do not sign anything that references release, arbitration, or waiver in exchange for a quick payment.
7. Contact a lawyer fast so a preservation demand reaches the defendant before logs cycle out.

Frequently Asked Questions

Can a company in California blame its AI system for injuring someone?

No. Under AB 316, a defendant that developed, modified or used an artificial intelligence system cannot escape liability by asserting that the AI acted autonomously. The company is judged under ordinary California negligence and product liability standards, exactly as if a human employee had made the decision.

Does AB 316 create a new right to sue?

It is a defense-limiting statute, not a new cause of action. You still sue under existing theories — negligence, strict product liability, failure to warn, negligent supervision, common carrier liability. AB 316 simply closes one specific escape route.

How long do I have to file an AI-related injury claim?

Generally two years from the injury under Code of Civil Procedure § 335.1. If a public entity is involved, a government claim is generally due within six months under Government Code § 911.2. Medical negligence and claims involving minors follow different clocks.

Who can be sued after an AI-caused injury in Los Angeles?

The deploying company, the model developer, the hardware manufacturer, the vendor or licensor, the employer that relied on the output, the property owner, and any maintenance or monitoring contractor. LA cases commonly name several at once.

What evidence matters most?

System and sensor logs, model version history, training and validation records, internal safety testing, prior incident reports, human-override data, and internal communications about known failure modes. Much of it is on short retention, so preservation cannot wait.

Does AB 316 apply to self-driving cars?

Yes — autonomous driving systems are squarely within its scope. Those claims also involve DMV autonomous vehicle regulations and standard California vehicle negligence law.

What if I was hurt at work by an automated system?

You likely have both a workers’ compensation claim against your employer and a separate third-party lawsuit against the robotics manufacturer, integrator or maintenance contractor. See our comparison of workers’ comp versus a personal injury lawsuit.

What does it cost to hire InjuryAllies?

Nothing upfront. We work on contingency — no fee unless we recover for you. Consultations are free. Call (323) 372-1216.

The Bottom Line

AB 316 reflects a straightforward judgment by the California Legislature: a company that chooses to put an automated decision-maker into the world accepts responsibility for what that decision-maker does. Delegating a task to software does not delegate the duty of care. For injured Angelenos, that removes what would have been the defense industry’s most attractive new argument, and it keeps AI cases on the well-mapped terrain of California negligence and product liability.

The practical challenge in these cases is no longer legal — it is evidentiary. The proof lives inside the defendant’s servers, and it does not stay there forever.