On November 25, 2025, the New York City Council passed three AI bills. The mayor returned them unsigned, and they became law on December 25, 2025. Their sponsors called the package the GUARD Act.
Local Law 188 creates an Office of Algorithmic Accountability inside the city’s technology department, led by a director the mayor appoints. The office must assess the risks of algorithmic tools before agencies deploy them. It reviews agency tools for risks of discriminatory decisions and for compliance, at least every four years or after design changes. It runs a public comment platform and a complaint process. It can require agency training, or suspend a tool that fails the city’s standards or poses a risk of biased decisions. It reports a summary of its assessments to the mayor and council by March 31 each year. The office’s charter section took effect June 23, 2026.
Local Law 193 has the office set basic compliance standards for “public-impacting” AI: AI that could materially affect people’s rights, benefits, safety, or access to city services. The standards apply when agencies develop, procure, deploy, or use such AI. They cover fairness, transparency, risk, monitoring, privacy, and verification of AI an agency did not build or control. They must be reviewed at least every other year.
Local Law 195 requires a public list of the systems the office has assessed. Each entry names the vendors, the purpose, the data the system can reach, and the risks found with recommendations. The first list is due March 31, 2027. Two limits: an agency can use a system that is not on the list, and the law makes the office’s recommendations advisory.
As of October 2026, we found no public announcement that the office is staffed or has completed assessments. A state audit in August 2026 found the city’s existing AI oversight only partly implemented.
The suspension power is the part with teeth. The public list is the part that lets anyone check.
Why it matters
City AI governance often starts as guidance. New York put an office with the power to suspend a tool into its charter, and tied it to standards that reach procurement.
For vendors, the laws change what a sale involves. A public-impacting system may need a pre-deployment assessment, verification by an agency that does not control the code, and an entry on a public list that names the vendor and the data the system can reach.
The gaps are about follow-through. The office depends on a mayoral appointment and funding. The list is not a gate. And the first public accounting is not due until March 2027.
Questions for agencies and vendors
- Who in your government can suspend an AI tool, and on what grounds?
- Which of your systems would count as public-impacting?
- Can a vendor support an assessment and compliance checks on a system the agency does not control?
- Are vendors prepared to have their name, purpose, and data access on a public list?
- Where does a resident go to question an algorithmic decision?
A law can create an office. Whether the office opens, staffs up, and uses its suspension power is the next record to watch.
Sources
- Int. 0199-2024, Office of algorithmic accountability, enacted as Local Law 188 of 2025 (New York City Council Legistar)
- Local Law 195 of 2025, list of assessed AI systems (intro.nyc)
- NYC Charter Section 20-u, Office of algorithmic accountability (American Legal Publishing)
- City Council establishes new AI oversight office (New York City Council, Nov. 25, 2025)
- State audit finds room for improvement in NYC’s AI governance (City & State New York, Aug. 2026)