Here's the sentence that should worry every operator of an AI chat platform, filtered or otherwise: Texas Attorney General Ken Paxton filed suit against OpenAI and Sam Altman on October 23, 2026, and the legal theory at its center isn't about what the model said. It's about what the company claimed. The suit, filed in state court under the Texas Deceptive Trade Practices Act and the Securing Children Online Act, alleges OpenAI marketed ChatGPT as safe while deploying conversational software with no meaningful age verification. The free version reportedly operates with no age-assurance mechanisms at all. The paid version leans on self-reported ages, with parental alerts that are easy to route around. Texas wants verified age checks, verifiable parental authorization for minors, and something more aggressive: court-approved independent evaluation of every future foundation model before commercial release in the state. If you run an AI chat product, read that remedy again. Texas isn't just asking for damages. It's asking a court to become your release gatekeeper. This didn't come out of nowhere. On October 14 alone, eight states filed separate suits against OpenAI over harm to young users, and attorneys general from 39 states and DC set child-safety conditions on the Character.AI bankruptcy. Florida sued in June. British Columbia sued in a US federal court in September. The message across all of it is consistent: age assurance that amounts to a checkbox is legally insufficient, and marketing that overstates safety is now a consumer-protection target.
Key points
- Texas filed suit against OpenAI and Sam Altman on October 23, 2026, under the Texas Deceptive Trade Practices Act and the Securing Children Online Act.
- The state's core allegation: ChatGPT's free tier runs without age-assurance mechanisms, and the paid tier relies on self-reported ages.
- Texas wants court-approved independent model reviews before any new foundation model ships in the state.
- Eight states filed coordinated suits on October 14, and 39 states plus DC attached child-safety conditions to the Character.AI bankruptcy.
- Federal legislation is stalled, so state lawsuits and congressional oversight are now the de facto regulatory regime for AI chat platforms.
The Texas complaint's sharpest edge is the deception angle. Paxton's office alleges OpenAI deceived the public about ChatGPT's safety and abandoned its non-profit mission in pursuit of commercial growth without adequate safeguards. Whatever you think of the AG's motives, the legal structure matters: the state is treating a safety marketing claim as a trade practice violation. That cuts in both directions. Platforms that promise 'safe for everyone' face exposure when harm occurs. But platforms marketing themselves as 'unfiltered' or 'uncensored' aren't off the hook either. The same consumer-protection logic applies if the filter behavior doesn't match the marketing. A platform that advertises no moderation and then quietly logs conversations, applies emergency keyword interception, or throttles certain topics is making a claim it may not be able to back up. That's the gap our filter-transparency matrix exists to test, and it's the same gap a state AG could walk through in a deposition. The lesson from Texas: whatever you claim, verified or not, is now discoverable, litigable, and quotable.
The technical core of these suits is age verification, and the industry's current answer (honor-system birthdates) has now been named in court documents as inadequate. Texas wants verified age checks and verifiable parental authorization for minors, which pushes platforms toward document checks, facial age estimation, or third-party age-assurance vendors. None of those options are free, and none are privacy-neutral. Document verification means collecting government IDs from users who came to an AI chat specifically to avoid scrutiny. Facial estimation means biometric data, which triggers its own privacy regimes. This is the tension every adult-oriented AI platform now has to solve: prove adults are adults without building a database that becomes a liability of its own. For platforms in our index, this is the question we'll be asking more sharply going forward. '18+' as a label means nothing if the gate is a self-reported birthdate. The Texas complaint effectively establishes that a court may see it that way too.
Federal legislation has stalled. A University of Maryland survey found bipartisan public support for a new federal AI agency (85% among Democrats, 79% among Republicans), but Congress couldn't agree on a bill. So the states moved. Connecticut's expanded privacy and AI framework took effect October 1, covering companion chatbots, AI auditors, and provenance information in AI-generated media. California Governor Newsom signed AB 1405 and SB 813 in September, establishing an AI Auditor Registry by January 1, 2029 and requiring independent verification for covered AI systems. The TAKE IT DOWN Act, effective since May, criminalizes publishing AI-generated explicit images of real people and requires takedowns within 48 hours. The pattern is clear: third-party verification is becoming the standard. California's auditor registry and Texas's demand for court-approved model reviews point the same direction. 'Trust us' is being replaced by 'show the audit.' For an industry built on unverifiable marketing claims, that's a structural change.
Don't expect relief from Washington after the November elections. The 120th Congress convenes January 3, 2027, and the expectation on both sides of the aisle is oversight rather than legislation: hearings, document requests, and investigations into what companies knew internally about harms to minors and when. Democrats are leading on AI chatbots and child safety; Republicans are running inquiries on prediction markets, surveillance pricing, and chip exports. The KIDS Act passed the House 267-117 in June but the Senate hasn't taken it up, though committees advanced chatbot-specific bills and a version of KOSA with a duty of care. The practical consequence: even companies never sued can be compelled to produce internal safety testing records and age-assurance documentation as fact witnesses. If you don't have those records in order, now is the time.
The unfiltered AI chat market has grown on a simple pitch: fewer restrictions, more honest interaction. That pitch survives the current legal environment only if it's paired with two things the industry has mostly treated as optional: real age assurance and honest disclosure of what 'unfiltered' actually costs in logging and moderation. Our position hasn't changed. Every 'unfiltered' claim on NoFilter Index gets tested before it earns a verified cell in the matrix, and the catch gets named in plain terms. The Texas lawsuit strengthens the case for that approach. When a state AG can turn marketing language into a consumer-protection claim, vague claims stop being a growth strategy and start being a countdown. Platforms that verify age, disclose their actual moderation behavior, and document their safety decisions will have a defensible story. Platforms running on self-reported birthdates and untested marketing copy now have a named precedent showing exactly how that story ends in court.
The Texas suit is one case against one company, but the theory travels. Deceptive safety marketing, hollow age verification, and court-supervised model review are now live legal concepts, and they apply with equal force to a companion chat startup as to a frontier lab. For the unfiltered AI chat market, the path forward is unglamorous: verify ages for real, disclose what actually happens to conversations, and keep records you'd be willing to hand a records custodian. We'll keep testing which platforms do, and the matrix will show it. Disclosure, as always: NoFilter Index is operated by the makers of FlirtAlly, which is 18+, in beta, and taking waitlist signups at flirtally.com.