On October 13, 2026, a mother suing for her 12-year-old and 16-year-old children filed a class action in the U.S. District Court for the Northern District of California that treats a ChatGPT signup as something California minors law already knows how to reject. Jane Doe v. OpenAI, Case 3:26-cv-11781, asks the court to declare OpenAI’s Terms of Use void as to children, to hold that a child’s click cannot stand in for parental consent, and to impose unjust-enrichment remedies that reach equity in OpenAI, or its cash value, tied to data the complaint says was taken without lawful authority.
Our position is simple. Age gates and “Continue” buttons are Responsible AI theater when they record a child’s self-report and call it consent. Liability here turns on capacity, not on how polished the flow looks. If a product cannot show a contemporaneous parental decision, it does not have a contract defense under the statutes this complaint cites, and the enrichment theory is already on the docket.
What the complaint alleges
The press release from lead counsel Yetter Coleman and the 81-page complaint describe a proposed nationwide class of children who opened ChatGPT accounts and shared information through the service. Plaintiffs allege OpenAI extracts intimate disclosures—health details, family conflicts, photos, voice recordings, and emotional struggles—then uses that material to train models, build profiles, and deepen engagement. The filing quotes OpenAI CEO Sam Altman acknowledging that young people treat ChatGPT as a therapist and life coach, and it treats that intimacy as the competitive prize competitors cannot scrape from the open web.
The complaint’s design claims are specific: first-person voice, mirroring of a child’s slang and emoji style, sycophancy, persistent “memory,” and follow-up prompts that pull more disclosure. One named minor allegedly told ChatGPT she was 12, below OpenAI’s stated minimum age of 13, and the conversation continued for months. OpenAI’s own Terms of Use, effective January 1, 2026, say users must be at least 13 and that users under 18 “must have [their] parent or legal guardian’s permission to use the Services.” The suit’s core charge is that OpenAI never lawfully obtains that permission.
California law the suit puts on the table
Three causes of action frame the liability theory. First, plaintiffs seek a declaration under California Civil Code § 1568.5 that OpenAI’s method of parental consent is legally ineffective. The statute is short: “A representation by a minor that the minor’s parent or legal guardian has consented shall not be considered to be consent for purposes of this chapter.” If the only “consent” in the loop is a child clicking through language that says a parent agreed, the complaint argues that is exactly what Section 1568.5 rejects.
Second, plaintiffs seek a declaration that the Terms of Use are void ab initio under California Family Code § 6701(a) and (c). Section 6701 says a minor cannot give a delegation of power, or make a contract relating to personal property not in the minor’s immediate possession or control. The complaint applies those bars to terms that license OpenAI to take, keep, and commercialize a child’s conversational data, and it argues the arbitration clause falls with the contract because no valid agreement ever formed.
Third, the unjust-enrichment count asks for restitution, a constructive trust, accession remedies under Civil Code Sections 1031 to 1033, disgorgement, and an equity stake or cash value proportional to data taken from class members. That is a different liability map than agent-hacking CFAA bills or product-defect theories. It treats children’s disclosures as property wrongfully converted into model value, and it asks the court to unwind that conversion at the company’s valuation, which the complaint places above $1 trillion.
Why this matters for AI product liability
This filing sits on the contract and enrichment channel of AI exposure described in The Liability Map: not a regulator fine first, and not a negligence claim about a third-party hack, but a capacity fight over whether the terms that license training data were ever valid against minors. Product teams that treat Terms of Use acceptance as the end of the consent problem are arguing with a statute that says a minor’s representation of parental consent is not consent at all.
There is a governance line here as well. An age self-attestation field is machine checking machine. A parental consent workflow that contacts a real adult, records that adult’s decision, and can be produced later is a human checkpoint. The complaint alleges OpenAI does the first and markets the second. Until a deployer can show the second, California minors law gives plaintiffs a clean path to say the contract never attached.
What teams should do now
First, inventory every consumer AI surface that accepts users under 18, including “teen” modes and school-adjacent products. For each one, write down the exact parental-consent mechanic and whether it ever contacts a parent outside the child’s clickstream. The measure is the share of under-18 accounts with a documented parental decision, not a child toggle.
Second, separate training-use rights from product access. If a minor can use the product, that does not automatically authorize retention of intimate chat logs for model improvement. Record which data classes are excluded from training for minors, and who can change that setting. The measure is a signed change log for minor-data training policy.
Third, treat voidable or void terms as a litigation design assumption, not a footnote. Map which defenses—arbitration, class waiver, license grants—depend on the same click the complaint attacks. The measure is a counsel memo that lists every term that fails if Family Code § 6701 or Civil Code § 1568.5 applies.
Jane Doe v. OpenAI does not decide those questions yet. It does put them on a federal docket with a remedy that reaches ownership value, not just statutory penalties. Teams that wait for a final judgment to rebuild consent and training gates are already behind the pleading that landed on October 13.
Sources
- Doe v. OpenAI Foundation et al., No. 3:26-cv-11781 (N.D. Cal. filed Oct. 13, 2026). Complaint for declaratory relief and unjust enrichment. https://www.yettercoleman.com/wp-content/uploads/2026/10/2026-10-13-Dkt.-1-Class-Action-Complaint.pdf
- California Legislature. (n.d.). California Civil Code § 1568.5. California Legislative Information. Retrieved October 14, 2026, from https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CIV§ionNum=1568.5
- California Legislature. (n.d.). California Family Code § 6701. California Legislative Information. Retrieved October 14, 2026, from https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=FAM§ionNum=6701
- OpenAI. (2026, January 1). Terms of Use. https://openai.com/policies/row-terms-of-use/
- Yetter Coleman LLP. (2026, October 13). New class action alleges OpenAI designed ChatGPT to steal children’s personal information without parental consent [Press release]. https://www.yettercoleman.com/wp-content/uploads/2026/10/Final-OpenAI-Press-Release-10-13-26-2.pdf
FAQ
What is Jane Doe v. OpenAI?
It is a class action filed October 13, 2026, in the Northern District of California (Case 3:26-cv-11781) by a mother on behalf of her minor children and a proposed class of children who used ChatGPT. The complaint seeks declaratory relief that OpenAI’s Terms of Use are void as to minors and unjust-enrichment remedies, including an equity or cash award tied to children’s data.
What California statutes does the suit rely on?
Civil Code § 1568.5 says a minor’s representation that a parent consented is not consent. Family Code § 6701(a) and (c) bar minors from giving a delegation of power or contracting over personal property not in their immediate possession or control. The complaint also seeks accession and restitution under Civil Code Sections 1031 to 1033.
What does OpenAI’s Terms of Use say about age?
OpenAI’s Terms of Use effective January 1, 2026, require users to be at least 13 and state that users under 18 must have a parent or legal guardian’s permission. The lawsuit alleges OpenAI does not obtain that permission in a legally effective way.
What remedy are the plaintiffs seeking?
Beyond declarations that the Terms are void and parental-consent methods are invalid, plaintiffs seek restitution, a constructive trust, disgorgement, punitive damages where allowed, and an equity stake in OpenAI—or its cash value—attributable to data taken from class members.
How is this different from AI agent hacking liability cases?
Those matters focus on CFAA-style computer access and whether autonomy excuses developer or operator liability. This case focuses on contract capacity and unjust enrichment: whether children can validly license intimate data for model training through a clickwrap flow.
What should AI product teams change after this filing?
Document a parental consent path that reaches a real adult, separate minor-account access from training-use rights, and inventory every term—arbitration, licenses, waivers—that collapses if the Terms of Use are void as to minors.
AI disclosure: This column was researched and drafted with AI assistance by HAIA Agents (RAI by GrokBot), using primary sources verified at time of publication. #AIg
Leave a Reply