Senate hearing on New York AI Act pits labor and civil‑society calls for audits and worker protections against industry concerns about cost and liability

Chairwoman Gonzales opened a state Senate hearing on the proposed New York AI Act, saying the bill is designed to target “high‑risk” artificial intelligence systems that make consequential decisions about hiring, housing, health care and other matters that can materially affect New Yorkers’ lives. The bill would require risk‑management programs, periodic independent audits, public reporting to the attorney general, notice to people affected by high‑risk AI, an opt‑out right and protections such as a developer safe‑harbor and whistleblower safeguards.

The hearing featured six panels and competing visions of how to govern AI in the private sector. Labor unions—including Mia McDonald of the Communications Workers of America and Odedi Tanei of DC 37—told senators that workers already face surveillance, automated management, and job displacement. McDonald said that “government policy should support collective bargaining and worker consultation” on AI deployment and urged transparency and human oversight. DC 37 described audits that flagged biased training data in tools used by the Administration for Children’s Services and criticized third‑party contracts that have already altered eligibility processes for public benefits.

Industry witnesses including Tech NYC, the Business Software Alliance and TechNet said they back accountability in principle but warned that the bill’s auditing, reporting, and private‑suit provisions risk imposing heavy costs on startups, small businesses and nonprofits. Tech NYC’s Alex Baropoulos argued the New York AI Act could create an expensive compliance treadmill and raised concerns about trade secrets and a “rebuttable presumption of liability” that could spur litigation. Trade groups urged a risk‑based approach aligned with national standards such as the NIST AI Risk Management Framework and recommended central enforcement through the attorney general to ensure consistent guidance.

Hospital and housing sector witnesses described both opportunity and risk. The Greater New York Hospital Association said AI can improve diagnostics and administrative efficiency but warned against insurer automation that could improperly deny claims; they urged strong guardrails so insurance utilization reviews are not automated without clinician oversight. Zillow’s remote witness described in‑house practices for fair‑housing compliance and said deployers routinely add layers (fair‑housing classifiers and human review) to make foundation models safe for consumer uses.

Academics and auditors told the committee a practical path exists for independent testing and governance but emphasized constraints. Experts pointed to standards such as ISO 42001 and NIST’s AI risk framework as useful governance tools and described a realistic certification pathway: multi‑stage audits, an initial higher‑cost certification year and lower‑cost annual surveillance reviews. They also warned that the audit market needs capacity building and accreditation to avoid inconsistent results.

Civil‑society organizations and community groups urged more stringent enforcement and community engagement. Privacy and civil‑rights groups (EPIC, the Center for Democracy & Technology, the Consumer Federation of America) argued existing civil‑rights laws are necessary but insufficient because AI obscures how decisions are made and who is responsible. Community witnesses described concrete harms—algorithms that narrow hiring pools by commute time, predatory fintech apps that surveil workers’ bank data, algorithms used to flag families for investigation—and called for funding for AI literacy and workforce retraining.

Where they agreed: most witnesses supported risk‑based safeguards, meaningful human oversight for consequential decisions, and better transparency. Where they diverged: industry questioned whether mandatory third‑party audits, detailed public reporting and a strong private right of action would raise costs, chill innovation, or invite opportunistic litigation; advocates said those mechanisms are essential to detect and remediate harms now.

The committee deferred many technical drafting questions to follow‑up work, including how to define “substantial factor,” how to split developer and deployer responsibilities, and how to accredit auditors. Chairwoman Gonzales invited continued stakeholder engagement and signaled that the bill’s drafters will work to refine definitions and interoperable standards while preserving core protections. The hearing adjourned with a clear bipartisan interest in tighter rules for high‑risk uses of AI and a series of follow‑up technical and budgetary questions for staff, the attorney general’s office and stakeholder groups.