AI liability at a glance
AI liability is the question of who pays, or answers to a regulator, when an automated system causes a legal harm. The answer so far comes from old law applied to new tools: employment discrimination statutes, consumer protection and unfair practices law, insurance contracts, securities fraud rules and product liability. No U.S. statute assigns liability to an AI system itself, and courts have not treated "the algorithm did it" as a defense.
The table below gives one line per area with its leading case or rule and the newest dated development. Each area has its own guide in the list on this page, and the sources for each line are linked in the sections that follow.
| Area | Leading case or rule | Status | Newest development |
|---|---|---|---|
| AI hiring | Mobley v. Workday, N.D. Cal. | Discovery; class motion pending | Class certification motion filed Sept. 14, 2026 |
| Hiring statutes | California SB 947; Colorado AI Act | Enacted, not yet in force | SB 947 signed Sept. 30, 2026 |
| Surveillance pricing | Maryland H.B. 895; New York disclosure law | Maryland ban in force | FTC 6(b) orders, Oct. 2026 |
| Insurance claim denials | Estate of Lokken v. UnitedHealth, D. Minn. | Discovery | Order compelling nH Predict records, Mar. 2026 |
| AI washing | SEC v. Delphia and Global Predictions; FTC Operation AI Comply | Settled actions | FTC sweep, Sept. 2024 |
| Chatbot product liability | Garcia v. Character Technologies; ChatGPT JCCP 5431 | Garcia settled; ChatGPT cases pending | Garcia dismissed after settlement, Jan. 2026 |
Is it legal for AI to reject a job application?
Yes, but the rejection has to comply with the same anti-discrimination laws that govern a human recruiter. If a screening tool rejects older, disabled, Black or female applicants at higher rates and the employer cannot justify the criteria as job-related, the outcome can be unlawful disparate impact even without intent.
Lawsuits
Mobley v. Workday, No. 3:23-cv-00770 (N.D. Cal.), is the leading case. In May 2025 the court conditionally certified a nationwide age-discrimination collective of applicants over 40, and in June 2026 it let California-law claims proceed because Workday runs its screening from California, as Duane Morris reported. The theory that a vendor can be liable as the employer's agent survived dismissal in 2024 and is the reason employers cannot treat a vendor contract as a shield. The full history is on the case tracker and in the Mobley case page listed below.
Hiring laws
States are adding process rules. California's SB 947, the No Robo Bosses Act, signed September 30, 2026, bars relying solely on an automated system to discipline or fire an employee and requires human corroboration and notice from July 1, 2027; California's AB 1883 restricts emotion-inference workplace surveillance from January 1, 2027; and the Colorado AI Act takes effect January 1, 2027, according to Gunderson Dettmer's fall 2026 employment update. Governor Newsom flagged missing definitions in the California bills when he signed them, so amendments are possible before they take effect.
Algorithmic and surveillance pricing
Surveillance pricing means setting a price for a person or group from their personal data. It is legal in most states if it is disclosed and not discriminatory. Maryland's Protection from Predatory Pricing Act, the first ban, took effect October 1, 2026 for food retailers and delivery services; Connecticut's ban takes effect July 1, 2027 and New Jersey's on August 1, 2027; and New York, which already requires an algorithmic pricing label, passed a broader ban in June 2026 that awaits the governor, according to Sidley Austin's October 2026 survey.
Federal enforcement runs through the FTC's power over deceptive and unfair practices. In October 2026 the FTC ordered Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart to explain how they set prices and gig worker pay, per the FTC's release. Those orders are a study, not an accusation. The surveillance pricing guide in this section covers the state laws in detail.
AI insurance claim denials
The main test case is Estate of Lokken v. UnitedHealth Group, No. 0:23-cv-03514 (D. Minn.), a proposed class action by Medicare Advantage members who say UnitedHealth used an algorithm called nH Predict, from its naviHealth unit, to cut off post-acute care that their doctors recommended. The plaintiffs allege that over 90 percent of the denials appealed were reversed. UnitedHealth denies that nH Predict makes coverage decisions and says physicians do.
In 2025 the court dismissed claims it found preempted by the Medicare Act but let breach of contract and good faith claims proceed, reasoning that the policy documents promised decisions by clinical staff and physicians, at 766 F. Supp. 3d 835. A March 2026 order then compelled UnitedHealth to produce documents on how nH Predict works and whether it was designed to replace physician judgment, according to Tressler LLP's case review. Tressler points to similar suits against Humana in the Western District of Kentucky and Cigna in the Eastern District of California. The contract theory matters beyond health insurance: if a policy says a person decides, an insurer that lets software decide may be in breach.
AI washing and FTC enforcement
AI washing is claiming more or different AI capability than a product has. It is treated as ordinary fraud or deception. On March 18, 2024, the SEC settled charges against two investment advisers, Delphia and Global Predictions, for false statements about their use of AI, with $400,000 in combined civil penalties; Global Predictions had called itself the first regulated AI financial advisor, per the SEC's announcement.
The FTC announced Operation AI Comply in September 2024 with five actions against companies that used AI claims to deceive consumers, including a service marketed as an "AI lawyer" and a tool for generating fake reviews, according to the FTC's release. The consumer side of AI washing also shows up in private suits: the Apple Siri AI settlement resolved false advertising claims over AI features that were promoted before they shipped.
Can you sue a company over an AI decision? Product liability and Section 230
Yes, and the first chatbot harm cases have tested how. In Garcia v. Character Technologies, a Florida mother sued after her 14-year-old son died by suicide following long conversations with a Character.AI chatbot. Judge Anne Conway let most claims proceed against the company, its founders and Google, and in January 2026 dismissed the case after the parties settled, as part of settlements in five cases reported by The Washington Post via Minnesota Lawyer. The settlements left open whether AI companies are liable for their models' outputs.
Wrongful-death and injury suits against OpenAI, including Raine v. OpenAI, are coordinated in San Francisco Superior Court as JCCP No. 5431, with no trial date, according to AI Lawsuit Tracker; the case tracker follows them. Section 230 of the Communications Decency Act protects platforms from liability for content posted by others; whether it covers text a company's own model generates is unresolved, and defendants have so far relied more on causation and First Amendment arguments. If you or someone you know is in crisis, call or text 988 in the U.S. to reach the 988 Suicide & Crisis Lifeline.
What laws protect consumers from AI?
No single federal AI consumer law exists. Protection comes from a layer of general rules: Section 5 of the FTC Act against deceptive and unfair practices, state consumer protection acts, civil rights and fair lending laws against discriminatory outcomes, and state privacy laws that give rights to opt out of data sales and some automated decisions. On top of these sit the new targeted statutes covered above: surveillance pricing bans, New York's pricing label, and California and Colorado rules on automated employment decisions.
For a consumer, the practical rights are concrete. You can ask whether a decision was automated where state law requires notice, appeal insurance denials (the Lokken plaintiffs allege most appealed denials were reversed), report deceptive AI claims to the FTC or a state attorney general, and file discrimination charges with the EEOC or a state agency within the deadline. The rights hub collects these routes, and the compliance guides cover what businesses must do.
Newest cases
- October 2026: FTC 6(b) orders on personalized pricing and pay to nine airlines, ride-share, delivery and retail companies.
- September 30, 2026: California SB 947 and AB 1883 signed.
- September 14, 2026: Rule 23 class certification motion filed in Mobley v. Workday.
- August 19, 2026: FTC proposed enforcement policy statement on personalized pricing.
- March 2026: Discovery order in Lokken v. UnitedHealth on nH Predict.
Each entry links through to its guide in the list on this page once that guide is live. Dates are taken from the sources cited in the sections above, and this list is updated as new rulings arrive.
When AI screens a job applicant, sets a price or denies a claim, the business using it usually remains liable, and courts are extending some liability to vendors. Workday faces a class certification hearing on March 9, 2027 in the age bias suit, Maryland already bans some surveillance pricing, and UnitedHealth is in discovery over algorithmic claim denials. The SEC and FTC treat false AI claims as fraud.
What changed
- 2026-11-09: Pillar published with status by topic.
Frequently asked questions
Who is liable when AI makes a mistake?
Usually the business that chose to use the AI, under the same laws that would apply if a person made the decision. Courts are also letting some claims proceed against vendors, as in Mobley v. Workday, where the software company is sued as the employers' agent.
Can you sue a company over an AI decision?
Yes. People have sued over AI hiring screens, insurance denials, chatbot harms and misleading AI marketing, using discrimination, contract, consumer protection and product liability law. Several cases have survived dismissal, and some have settled, but few have reached a final judgment on liability.
What laws protect consumers from AI?
Mostly general laws: the FTC Act, state consumer protection and civil rights laws, insurance contracts and privacy statutes. Newer targeted laws include Maryland's surveillance pricing ban, New York's algorithmic pricing label, and California and Colorado rules on automated employment decisions taking effect in 2027.
Is it legal for AI to reject a job application?
Yes, if the screening does not discriminate. An AI tool that rejects protected groups at higher rates without a job-related justification can violate federal and state anti-discrimination law, and California will require human review of some automated employment decisions from July 1, 2027.
Sources
- Duane Morris: Mobley v. Workday motion to dismiss ruling (June 24, 2026)
- Gunderson Dettmer: quarterly employment law update, fall 2026
- Sidley Austin: state personalized pricing laws (Oct. 26, 2026)
- FTC: personalized pricing and compensation inquiry (Oct. 2026)
- Tressler LLP: Estate of Lokken v. UnitedHealth Group (June 1, 2026)
- SEC press release 2024-36: charges against two advisers for AI washing
- FTC: Operation AI Comply announcement (Sept. 2024)
- Minnesota Lawyer (Washington Post): Google, Character.AI move to settle suits (Jan. 9, 2026)
This page is journalism and commentary, not legal advice. AI law changes quickly; check the primary sources above and consult qualified counsel for your situation.