By Sami Sedghani, Pharm.D., J.D. · Synergist Law, P.C. · San Rafael, California · Updated October 24, 2026
This page provides general information about California law as it applies to injuries involving artificial intelligence. It is not legal advice, and reading it does not create an attorney–client relationship. This area of law is new and changing quickly, and every case depends on its own facts. For advice about your specific situation, consult a licensed California attorney.
When an AI system is involved in an injury, California law asks the same question it always has: who made a choice that caused harm? A company that builds an AI product, a business that decides to rely on one, and an insurer that lets an algorithm drive a coverage decision are all making choices. Since January 1, 2026, California has also closed off one argument these defendants might have tried. Under Civil Code § 1714.46, a defendant that developed, modified, or used artificial intelligence may not defend itself by claiming the AI caused the harm on its own.
Synergist Law is led by attorney Sami Sedghani, a trial lawyer and former clinical pharmacist in San Rafael. Before founding Synergist Law, Sami spent nine years at large firms defending companies in complex and product liability litigation. This page explains how California law applies when AI is part of what went wrong, and where these cases are being decided.
“The AI did it” is not a defense in California
Civil Code § 1714.46, added by AB 316, applies to any defendant that developed, modified, or used AI alleged to have caused harm. That defendant may not assert that the AI acted autonomously. The statute does not make anyone automatically liable. A defendant can still argue that its conduct did not cause the injury, that the harm was not foreseeable, or that someone else shares fault. What it cannot do is treat the software as a separate actor that absorbs the blame.
That matters most in the second kind of case described below, where a business uses an AI tool and the tool makes the mistake. The business chose the tool, decided how much to trust it, and decided whether a person would check its work. Those are ordinary negligence questions.
Two kinds of AI injury cases
1. Harm caused by an AI product itself
Here the claim runs against the company that designed and sold the product. The legal framework is California product liability (design defect, failure to warn) and negligence. The most visible examples today are conversational AI chatbots, where families have alleged in court that a chatbot encouraged self-harm or deepened a user’s delusions. Whether a chatbot is a “product” for strict liability purposes is being litigated now. In Garcia v. Character Technologies, a federal court in Florida allowed product liability claims to proceed in May 2025 (the case later settled), and OpenAI has described ChatGPT in court filings as a “software-based service” rather than a product.
Our page on AI chatbot harm, including risks to minors covers these cases in detail. Self-driving vehicles are another example, covered on our self-driving car and robotaxi accident page.
2. Harm caused by a business that relied on AI
Here the AI developer may not be the main defendant at all. The claim runs against the hospital, clinic, insurer, employer, or fleet operator that put an AI tool into a decision affecting someone’s safety. The questions are familiar: Was the tool reasonable to use for that task? Did the business test it, train its staff, and supervise its output? Did a qualified person review the decision before it reached the patient, driver, or customer? A business that hands a safety decision to software keeps its duty of care.
Where AI is showing up in injury cases
Conversational chatbots
Lawsuits alleging psychological harm, self-harm, and wrongful death have been filed against several chatbot makers. In California, a group of product liability cases against OpenAI was coordinated in February 2026 in San Francisco Superior Court as In re ChatGPT Product Liability Cases, JCCP No. 5431. A wrongful death case involving Google’s Gemini, Gavalas v. Google LLC, is pending in federal court in San Jose. These are allegations that have not been proven. See our AI chatbot harm page.
Health care
Hospitals and clinics now use AI to draft visit notes, flag abnormal results, triage messages, and suggest diagnoses. A clinician who relies on AI output is still held to the professional standard of care. If an AI-drafted note records the wrong medication or dose and no one catches it, or a flagging tool misses a finding a radiologist would have seen, the question is whether the providers acted reasonably in how they used the tool. Medication history is often where these errors surface, and it is where Sami’s pharmacy training is most useful. Medical malpractice claims in California carry their own notice requirements and shorter deadlines, so timing matters.
Insurance coverage decisions
California has addressed this directly. SB 1120, effective January 1, 2025, amended Health and Safety Code § 1367.01 and Insurance Code § 10123.135. Health plans and health insurers may not use an AI tool to deny, delay, or modify care based on medical necessity. A licensed physician or qualified health care professional must make that determination after reviewing the patient’s own records, and the tool cannot rely solely on group data. When a denial delays care and the delay causes harm, the plan’s use of automation can become part of the evidence. Claims against health plans have special rules: many employer plans are governed by federal ERISA law, and California generally requires independent medical review first.
Vehicles and robots
Driverless robotaxis, driver-assistance systems such as Tesla’s Full Self-Driving (Supervised), and sidewalk delivery robots are covered on our self-driving car page. Fleets that use AI to set routes and delivery windows raise a different question: whether the schedule the software produced pushed drivers into unsafe driving. That issue is covered on our commercial vehicle accident page.
The evidence is different in an AI case
- Logs and transcripts. Chat histories, prompts, outputs, and timestamps. For a chatbot user, the account itself is evidence and should not be deleted.
- Version history. Which model or software version was running on the date of the injury, and what changed before and after.
- Safety testing and internal warnings. What the company knew about the risk and when.
- Vendor contracts. When a business licensed an AI tool, the contract shows what the vendor promised, what the business agreed to monitor, and who carries the insurance.
- Human review records. Whether a person checked the AI output before it was acted on, and how often reviewers overrode it.
Much of this sits on company servers and can be changed or overwritten. A written preservation demand puts the company on notice early. If evidence is later lost or destroyed, California juries may be instructed that they can draw an adverse inference from willful suppression (CACI No. 204).
Why California, and why the Bay Area
Many of the most closely watched AI injury cases are being litigated in California courts: the coordinated ChatGPT cases in San Francisco Superior Court and the Gemini wrongful death case in the Northern District of California in San Jose. California has also passed some of the first statutes aimed at these products, including Civil Code § 1714.46, SB 1120, and the companion chatbot law, SB 243 (Business and Professions Code §§ 22601–22606). For an injured person anywhere, it is worth understanding how California courts are treating these claims. Synergist Law is in San Rafael, in Marin County, across the Golden Gate Bridge from the San Francisco courthouse, and handles AI injury matters for clients in Marin, the Bay Area, and across California.
Deadlines
- Personal injury and wrongful death: generally two years under Code of Civil Procedure § 335.1. For a minor’s own injury claim, the deadline is generally paused until the minor turns 18 (Code Civ. Proc. § 352). That pause does not apply to medical malpractice claims or claims against public entities, which have their own rules for minors.
- Medical malpractice: shorter limits and a pre-suit notice requirement apply, including when AI tools were part of the care.
- Public entities: a written government claim generally within six months under Government Code § 911.2, for example when a public hospital or agency used the tool.
- Terms of service: some AI products set their own notice deadlines or require arbitration. Read the terms early, and do not assume they control a minor’s claim.
How we approach these cases
Technology companies defend these cases on two fronts: the system (“it worked as designed” or “it is a service, not a product”) and the person (“the harm came from a pre-existing condition”). Sami spent nine years on the defense side of complex and product liability litigation, so we know how a corporate defendant builds both arguments and which internal records those defenses depend on.
The second front is where many AI cases will be won or lost. Psychiatric history, medications, and treatment records are central, and a defense expert will read them looking for an alternative cause. Because Sami is a former clinical pharmacist, we read those records the same way, before the defense does.
Neither background predicts how any particular case will turn out. It shapes how we prepare one.
Frequently asked questions
Can a company blame its AI for causing an injury in California?
Not as a defense. Since January 1, 2026, Civil Code § 1714.46 bars a defendant that developed, modified, or used AI from arguing that the AI autonomously caused the harm. The defendant can still contest causation, foreseeability, and comparative fault.
Who can be responsible when AI is involved in an injury?
Depending on the facts, the company that built the AI product, the business that chose to use it in a safety-related decision, and any person who should have reviewed its output. A business that relies on AI keeps its own duty of care.
Can a health insurer use AI to deny care in California?
Not for medical necessity decisions. Under SB 1120, effective January 1, 2025, a licensed physician or qualified health care professional must make medical necessity determinations after reviewing the patient’s own records, and AI tools cannot be used to deny, delay, or modify care on that basis.
Is an AI chatbot a product that can be defective?
Courts are deciding that now. A federal court in Florida allowed product liability claims against Character.AI to proceed in 2025 before that case settled, while OpenAI has described ChatGPT in court filings as a software-based service rather than a product. The answer may depend on which design features a claim targets.
How long do I have to bring an AI injury claim?
Most California personal injury and wrongful death claims must be filed within two years, and a minor’s own claim is generally paused until age 18, except for medical malpractice and public-entity claims. Medical malpractice and claims against public entities have shorter deadlines, and some terms of service add their own. Because the deadline depends on the facts, it is better to have a licensed California attorney determine the deadline that applies than to calculate it yourself.
Talk to a California AI injury lawyer
If you or a family member was harmed and an AI system was part of what happened, Synergist Law is willing to review it. Sami Sedghani, Pharm.D., J.D. (Super Lawyers 2022–2026; appellate counsel in the published decision Osaze v. Gee (2026)) handles these matters personally. Se habla español.
Tenemos una guía en español sobre casos de lesiones personales: Abogado de lesiones personales en California.
Office: 1299 4th Street, Suite 301, San Rafael, CA 94901. Phone: (415) 835-4448. 24/7 answering service. Contingency fee: no attorney’s fee unless we obtain a recovery for you. Depending on the terms of your written fee agreement, you may be responsible for case costs we advance even if there is no recovery. Your fee agreement explains this before representation begins. Request a free consultation or review our frequently asked questions.
Disclaimer: This page provides general information only and does not constitute legal advice. No attorney–client relationship is formed by reading this content. References to lawsuits describe allegations made in court filings, which have not been proven. Statutes, rulings, and company practices described here may change after the date shown above. Every case turns on its own facts, and nothing on this page predicts any outcome, timeline, or value in your matter. For advice specific to your situation, consult a licensed California attorney.