Mayor moves to kill unusable Adams-era system costing half million dollars annually as first efficiency measure
Mayor Zohran Mamdani identified the Eric Adams administration’s artificial intelligence chatbot as an early target for elimination, citing its dysfunction and cost as symptomatic of wasteful government spending. During a February 2026 press conference addressing the city’s 12 billion dollar budget deficit, Mamdani singled out the chatbot as an example of failed technology implementation and promised to identify similar inefficiencies throughout the municipal budget.
System Failures and Public Interest Reporting
The chatbot, released by the Adams administration in fall 2023 as part of its digital transformation initiative, was designed to provide business owners with accessible information about New York City regulations and compliance requirements. However, investigative reporting by The Markup and The City revealed that the system confidently provided illegal and harmful advice to users. Testing documented instances where the chatbot recommended actions that violated labor laws, housing discrimination statutes, and municipal ordinances. The system suggested landlords could refuse Section 8 housing vouchers despite federal fair housing protections, instructed users that refusing cash payments was legal contrary to city ordinance, and consistently failed to provide accurate information about minimum wage requirements.
Budgetary and Technical Implications
Construction of the chatbot’s foundation reportedly cost nearly 600,000 dollars, with Mamdani citing annual maintenance costs around 500,000 dollars. The mayor characterized the system as “functionally unusable” and positioned its elimination as indicative of broader problematic spending patterns under his predecessor. The decision reflects pragmatic budget management at a moment when the city faces substantial fiscal pressure.
Broader Technology Governance
According to The Markup reporting, the chatbot failure represents a cautionary tale about deploying artificial intelligence in government without adequate testing and legal vetting. The system’s inability to recognize its own limitations and its tendency to confidently assert incorrect information exemplifies challenges inherent in applying language models to high-stakes decision-making contexts. Experts noted that AI systems, however sophisticated, cannot reason through legal principles or apply contextual judgment required for regulatory compliance guidance. Mamdani’s decision signals his administration’s skepticism toward technological solutions disconnected from human expertise and accountability.
