Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Consumer Protection topic
No spam. Unsubscribe anytime.
Attorney General's Office urges disclosure, watermarking and audits at Assembly AI hearing
Summary
New York Attorney General's Office told an Assembly joint hearing that disclosure when consumers interact with AI, model and output auditing, and strengthening GBL 3—49 are practical first steps to protect consumers from bias, fraud and harmful AI outputs.
The New York Attorney General's Office told the Assembly's Consumer Affairs and Protection and Science and Technology committees that transparency and stronger enforcement are the foundation of consumer protection for artificial intelligence.
Chris D'Angelo, Chief Deputy for Economic Justice, and Darsana Srinivasan, Chief of the Healthcare Bureau, testified that generative and machine-learning AI bring "remarkable benefits" but also risks including targeted scams, biased decision-making and misleading advertising. "Consumers should know when they're interacting with AI content," D'Angelo said, urging disclosure obligations for commercial contexts and watermarking to enable provenance tracing.
The AG's team recommended three core measures: disclosure so people know they're dealing with AI-generated content; auditing for bias, discrimination and accuracy; and mandatory reporting of significant failures or "hallucinations." The office also called for an update to GBL 3—49 to add an "unfairness" prong that would give enforcers broader authority to address abusive or discriminatory AI-enabled practices. "Auditing, to improve outputs for decision making, and identification through watermarks and other tracking technology," the testimony said.
Committeemembers pressed the AG's office on enforcement mechanics. Darsana Srinivasan told legislators that both predeployment model review and post-deployment outcome auditing have roles to play: "Both have their important use cases and purposes," she said, adding that output audits are an accessible first step while model-level review can provide deeper assurances where risk is greatest. The office also said updates to consumer-protection statutes would allow existing tools to be applied to AI-driven fraud and deception.
The AG's witnesses noted resource constraints and the limits of civil authority for criminal harms, saying civil enforcement remains a principal tool but that coordination with federal partners would be used when criminal conduct is suspected. They urged lawmakers to adopt practical, enforceable obligations that improve the ability of regulators and the public to hold model creators and deployers accountable.
The committee did not take votes; chairs said testimony will be posted to the Assembly's website and staff will follow up on technical issues such as watermarking standards.
