Tech Industry Opposes ‘Broad’ Maryland AI Transparency Bill
An AI transparency bill introduced in the Maryland General Assembly is overly broad and anti-competitive, tech industry representatives told state lawmakers Tuesday.
Sign up for a free preview to unlock the rest of this article
Del. Chao Wu (D) introduced HB-823 on Jan. 29. He told the House Economic Matters Committee during a hearing Tuesday that Maryland should take note of AI transparency laws in Colorado and Utah. The proposed bill would require “high-level” summaries about datasets used to train generative AI systems to prevent discrimination and other inequities, its author said. HB-823 would require that developers publicly disclose on their websites the “data used to train” generative AI models.
The committee took testimony during first reading from Chamber of Progress, TechNet and the Computer & Communications Industry Association but didn’t vote on the proposal. CCIA State Director Megan Stokes said the legislation departs from global transparency standards and would force companies to publicly disclose AI-related trade secrets. She and TechNet Executive Director-Mid-Atlantic Region Margaret Durkin urged the committee to amend the bill and explicitly state there’s no private right of action.
Durkin said the dataset disclosures would overwhelm consumers due to the sheer volume of data. A better obligation would require that companies provide summaries of the data used in training AI systems, Durkin said. She suggested companies should only be required to retain data for auditing purposes, not to share it publicly on their websites.
Maryland Chamber of Commerce Vice President-Government Affairs Grayson Wiggins said he’s not sure any state has enacted an AI transparency bill requiring this level of disclosure.
Wu cited the Utah AI Policy Act and the Colorado AI Act as similar measures. Both require generative AI disclosures and neither includes a private right of action. Colorado’s law goes into effect in February 2026, and Utah’s went into effect in May.
Chamber of Progress Northeast State Director Briana January said the Maryland bill would require that companies disclose the “full recipe” for generative AI systems, not just the “nutritional label.” She cited broad disclosure requirements for training datasets, including the source and characteristics of the data. That would allow companies to reverse engineer their rivals’ AI models, she said.