Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Transparency topic
No spam. Unsubscribe anytime.
Generative AI transparency bill wins mixed response from industry and advocates
Summary
Delegate Chao Wu presented HB823 to require high‑level public summaries of the data used to train generative AI. Tech industry groups supported transparency in principle but urged narrower, non‑proprietary disclosures and trade‑secret protections; financial regulators sought alignment for banks already subject to oversight.
Delegate Chao Wu told the committee HB823 would require developers of generative AI systems to publish a high‑level summary of the datasets used to train their models, arguing transparency is needed to detect bias and to build public trust.
"By mandating the developers publish clear documentation about the training dataset, this bill ensured that bias in AI will not be identified and corrected before the systems are deployed to the public," the sponsor said.
Industry witnesses and trade associations endorsed the goal of transparency but asked the committee to narrow disclosure obligations to avoid revealing trade secrets or exposing proprietary model details that could enable reverse engineering or security vulnerabilities. Brianna January (Chamber of Progress) and Margaret Durkin (TechNet) recommended less granular, category‑level descriptions and suggested private reporting or audit rights instead of full public disclosure.
The Maryland Bankers Association urged carveouts or equivalence when financial institutions are already subject to regulator guidance, saying bank supervisors and examiners already review AI use and training materials as part of prudential supervision.
Opponents cautioned that overly prescriptive public disclosures could disadvantage small innovators and suggested the committee examine private disclosure to regulators, trade‑secret protections and an explicit prohibition on a private right of action. The bill drew cross‑sector interest and calls for a negotiated middle ground between consumer transparency and IP/security protections.
