AI is good and growing fast but we are still in the developmental stages. Companies need to be careful with the claims they make. Texas Strikes Settlement with Dallas AI Firm Over Misleading Healthcare Tech Claims 📢 Issue: The State of Texas, through its Attorney General, filed a petition for the approval and entry of an Assurance of Voluntary Compliance (AVC) with Pieces Technologies, Inc. The key issue involves whether Pieces Technologies, Inc. engaged in deceptive practices under the Texas Deceptive Trade Practices - Consumer Protection Act (DTPA) by making false or misleading representations about the accuracy of its artificial intelligence (AI) products used in healthcare. 📢 Rule: The Texas Deceptive Trade Practices Act (DTPA) prohibits false, misleading, or deceptive practices in the advertising and sale of goods or services (Tex. Bus. & Com. Code §§ 17.41-.63). Specifically, Section 17.47 allows the Consumer Protection Division of the Attorney General’s office to investigate and take action on violations. 📢 Application: Allegations by the State of Texas: Pieces Technologies developed and marketed AI products intended to assist healthcare providers in treating patients. It claimed its products had very low error rates (referred to as "hallucination rates") in creating outputs like clinical notes. The state alleged these claims were false or misleading, violating the DTPA, since the AI could produce incorrect or misleading outputs. Response by Pieces Technologies: The company denied any wrongdoing or liability, stating that its claims about the AI’s accuracy were accurate and did not violate any laws. Assurance of Voluntary Compliance: To settle the matter without prolonged litigation, Pieces Technologies agreed to a series of actions, including: Providing clear and conspicuous disclosures in all marketing materials regarding the accuracy and metrics of its AI products. Prohibiting misleading or unsubstantiated representations about the accuracy or testing of its AI products. Ensuring transparency with customers about the potential risks and limitations of its products. Complying with the AVC terms for five years and implementing internal processes to monitor compliance. 📢 Conclusion: Pieces Technologies agreed to settle the matter by entering into the AVC, which does not constitute an admission of liability. The company agreed to specific practices and disclosures to comply with Texas law, as outlined in the AVC, and the court’s approval was sought for the AVC. --------------- Sanjay Juneja, M.D. Douglas Flora, MD, LSSBB, FACCC Nixon Gwilt Law Sean Weiss Ronald Chapman II, Esq. LLM Michael Crocker Rebecca E. Gwilt David Penberthy, MD MBA FACCC VMG Health ECG Management Consultants Etyon HLTH Inc. Aashish Shah ------ https://lnkd.in/enduqyDe.
Sean Weiss
4mo
Jordan Johnson, MSHA, iMPaCT this is our topic for #thecomplianceguy podcast next week! Providers need to understand that AI for clinical notes at this point is nothing more than a tool and requires human engagement on the front and back ends! Patient disclaimers on the use of AI in their clinical notes is also important. I have also learned that there are scribe programs out there that will automatically add a diagnosis code, but what the providers fail to realize is that the diagnosis code is only the base code, and it has to be taken to the highest level of specificity to ensure the medical necessity of the claims being submitted. AI in healthcare compliance has a long way to go!
Wow! Thats very interesting and makes me wonder about the margin of error for other AI driven products and services used in healthcare. We are becoming more and more dependent on this tech because of the demanding nature of the business…. But AI is only as solid as its development and the information it’s fed.
Al Lawson
4mo
Thanks Jordan for bringing this forward!
Hematologist & Medical Oncologist | Host of 'AI and Healthcare' Podcast | Keynote | Editorial Board Member 'AI in Precision Oncology' | Co-Founder of Medfluencers & TensorBlack AI | Citrus Oncology
4moso glad you shared this! it's imperative to be humble, and cautious, when challenging generalizability. over fitness and--usually more of an ascertainable concern during training, hallucination rate--are known problems we need to respect if we hope to get productive use and utility from these AI tools !