Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Transparency topic
No spam. Unsubscribe anytime.
Supporters and tech groups debate AI training‑data disclosure in HB823; industry urges trade secret protections
Summary
HB823 would require developers of generative AI systems to publicly disclose high‑level information about training data; tech trade groups and industry coalitions supported transparency in principle but asked for narrower, non‑proprietary disclosures and trade‑secret protections; consumer advocates supported the bill with caveats.
House Bill 823 would require developers of generative artificial intelligence systems to publish high‑level summaries of the data used to train their models. Delegate Chao Wu, the bill sponsor, told the Economic Matters Committee the requirement aims to improve accountability, reduce bias, and foster public trust in AI systems used by Marylanders.
"Without transparency about data use, it becomes really impossible to identify or address issues like bias," Delegate Wu said. The bill would require public disclosures on company websites describing training datasets in broad categories rather than publishing raw data.
Industry response: Several tech trade associations — Chamber of Progress, TechNet, Computer & Communications Industry Association (CCIA) — and industry groups such as the Maryland Bankers Association supported transparency in concept but asked for significant changes. Chamber of Progress urged narrower disclosures that provide a "nutritional label"‑style summary rather than a step‑by‑step recipe, warning that overly detailed data disclosures could reveal proprietary methods and deter innovators, especially startups.
TechNet and CCIA argued the bill should allow trade‑secret or IP protections and permit certain disclosures to be made privately to state regulators instead of posted publicly. The Maryland Bankers Association asked for language recognizing existing federal and state banking regulator regimes, so banks can rely on those regulators' work and not face duplicate obligations.
Opposition and concerns: The Maryland Chamber of Commerce and the Consumer Data Industry Association emphasized costs and operational burdens on businesses, especially small startups, and the risk of overwhelming consumers with excessive technical detail. The Consumer Data Industry Association also argued the proposal could be preempted by federal law in some respects unless carefully calibrated.
Committee notes: Witnesses urged adding exemptions for trade secrets, clarifying definitions, avoiding private rights of action, and tailoring disclosure obligations to avoid competitive harm while still providing meaningful information. Delegate Wu and supporters said they would work with stakeholders on language to balance transparency and innovation.
Ending: The bill drew broad interest and substantive requests for amendments to protect trade secrets, narrow disclosure scope, and harmonize with banking and regulatory frameworks. No committee vote is recorded in the provided transcript excerpt.
