AI Regulation Tracker / Enforcement action
Texas Used an Old Consumer-Protection Law to Bring the First State AI Enforcement Case
On September 18, 2024, the Texas Attorney General settled with Pieces Technologies, a Dallas healthcare AI company, over deceptive claims about how accurate its hospital tool was. The instrument was not an AI statute. It was the Texas Deceptive Trade Practices Act, and that choice is why this case is the template every attorney general now has for going after AI accuracy marketing.
For most of the AI debate, the assumption has been that regulators need a new AI statute before they can act. Texas showed that assumption is wrong. The Attorney General did not reach for a bespoke AI law, because there was no operative one to reach for. He reached for the Deceptive Trade Practices Act, the same broad consumer-protection tool the office uses against any business that lies to the market. The product happened to be generative AI. The legal theory was ordinary deception.
The facts made the theory easy. Pieces sells a tool that reads a patient’s chart and writes a summary of that patient’s condition and treatment for the people caring for them. As the AG’s office put it, at least four major Texas hospitals were “providing their patients’ healthcare data in real time to Pieces so that its generative AI product can ‘summarize’ patients’ condition and treatment for hospital staff.” That is about as high-stakes as software gets. A wrong summary in that context is not a typo. It is a clinical risk.
The number that triggered the case
Pieces marketed its accuracy with a hard-looking metric. The company claimed a “severe hallucination rate” of “<1 per 100,000,” which is the same as saying an error rate under a thousandth of a percent. The AG’s investigation found those metrics were “likely inaccurate and may have deceived hospitals about the accuracy and safety of the company’s products.” This is the part every AI marketer should sit with. The problem was not that the product could hallucinate. Every generative system can. The problem was publishing a precise, reassuring number that the company could not stand behind. A specific accuracy claim is a factual representation, and a factual representation that is not true is deception under the DTPA.
What Pieces agreed to do
The settlement is a disclosure-and-transparency remedy, not a ban. Under it, Pieces agreed to accurately disclose the actual extent of its products’ accuracy rather than lean on a flattering metric. It also agreed, in the AG’s words, to “ensure that the hospital staff using its generative AI products to treat patients understand the extent to which they should or should not rely on its products.” In plain terms: state the real accuracy, state the intended use, and make the limitations and appropriate-use boundaries clear to the clinicians on the other end. The obligation runs to the buyer and to the humans making decisions with the output, which is exactly where the risk lives.
Paxton drew the line in one sentence
The Attorney General framed the standard directly. “AI companies offering products used in high-risk settings owe it to the public and to their clients to be transparent about their risks, limitations, and appropriate use. Anything short of that is irresponsible and unnecessarily puts Texans’ safety at risk,” Paxton said. He did not stop at the vendor. He put the buyer on notice too: “Hospitals and other healthcare entities must consider whether AI products are appropriate and train their employees accordingly.” Read those two lines together and you have the whole compliance posture. The vendor must be honest about limits. The buyer must do its own diligence and cannot outsource that judgment to a marketing metric.
Why this is the template, not a one-off
The durable lesson is the instrument choice. Every state has a consumer-protection statute like the DTPA, and every attorney general already knows how to run a deception case with it. That means AI accuracy marketing is exposed to enforcement in essentially every state right now, with no new legislation required and no waiting. The elements are familiar: a factual claim, made to a buyer, that is false or unsubstantiated, in a way that matters. A hallucination rate, an accuracy percentage, a “clinically validated” label, a benchmark score in a sales deck, each is a factual representation a general counsel can be asked to defend. This case is the proof of concept that regulators noticed and have been building on since.
What this means for professionals
If you market AI, treat every quantified accuracy claim as a legal representation and keep the substantiation that backs it, because the gap between the number in the deck and the number you can prove is the whole case. If you buy AI for a high-stakes function, Paxton’s second sentence is aimed at you: independently assess whether the tool is appropriate, get the real accuracy and limitation disclosures in writing, and train the people who will act on the output. If you are counsel, the practical move is to map your AI vendors’ public accuracy claims against what their own documentation supports, and to build the same disclosure obligations Pieces accepted into your contracts before a regulator builds them for you. The through-line regulators keep returning to holds here too: automation is allowed, but someone stays accountable for the truth of what was promised.
Questions professionals are asking
What law did Texas actually use?
The Texas Deceptive Trade Practices Act, a general consumer-protection statute, not an AI-specific law. The Attorney General treated Pieces’ accuracy marketing as a false or misleading representation, the same theory used against any business that deceives the market. The product being generative AI did not change the legal test.
What exactly did Pieces claim?
Pieces marketed its hospital generative AI product as highly accurate, including a “severe hallucination rate” of “<1 per 100,000.” The AG’s investigation found those metrics were likely inaccurate and may have deceived hospitals about the accuracy and safety of the products.
Is this binding law or just a settlement?
It is a settlement, binding on Pieces, that resolved the investigation without a court finding or admission of liability. It does not create new statutory law. Its importance is as precedent: it shows that existing consumer-protection statutes reach AI accuracy claims, which is why it functions as a template for other attorneys general.
What did Pieces have to do?
Accurately disclose the real extent of its products’ accuracy instead of relying on a flattering metric, and ensure hospital staff understand the extent to which they should or should not rely on the output. In practice that means honest accuracy disclosure plus clear communication of limitations and appropriate use to the clinicians using the tool.
Why does a 2024 Texas case matter to me now?
Because the mechanism is portable. Nearly every state has a consumer-protection statute like the DTPA, so any quantified AI accuracy claim is exposed to a deception action right now, with no new AI law required. If your company publishes accuracy or error-rate numbers, keep the substantiation, and if you buy AI for high-stakes work, get the real limitation disclosures in writing.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal or compliance advice. Confirm how the Texas Deceptive Trade Practices Act and AI accuracy-claim enforcement apply to your situation with qualified counsel.