In February 2026, New York’s counties announced a shared dashboard for rating the AI tools they use. Route Fifty reported it on February 26, after officials presented it the week before at the National Association of Counties’ legislative conference. The New York State Association of Counties (NYSAC) backs it with Computer Aid, Inc. (CAI), a services firm, as part of a program the two call the GovAI Trustmark.
Counties rate tools on a standardized 10-point scoring framework. The tools include ChatGPT, Google Gemini, Microsoft Copilot, and Claude. Each county can see the others’ evaluations and procurement documents, and evaluators are identified. The stated goal is to stop every county from vetting the same tool on its own, and to show whether peers actually use a tool and whether it works as intended.
The need came from the counties’ own survey. Fifty-seven percent said they lack guardrails for AI use, Route Fifty reported. A NYSAC and CAI survey found 63 percent of respondents, 188 of 293, actively use at least one generative AI tool. ChatGPT led, used by 79 percent of active users. Concerns included accuracy, confidentiality, and training.
The platform has since grown. In July 2026, CAI announced a statewide launch with 85 users across 32 of the state’s 62 counties, 20 tools available for evaluation, and 20 county evaluations completed. Those figures come from CAI’s press release. The 10 scoring criteria have not been published in the coverage we found. NYSAC describes them as covering cost, security, ease of use, risk, data protection, and how well tools work for users.
The counties are pooling the one thing a small IT shop cannot afford to repeat: due diligence.
Why it matters
A county with a small IT staff buys the same AI tools as a large one, without the capacity to evaluate them. Shared ratings and shared contract documents work like cooperative purchasing for judgment. One county’s review becomes another county’s starting point.
The limits are real. A peer rating is only as good as its criteria and its evaluator. Twenty evaluations is an early sample. And the platform is run with a vendor partner, so counties should know who owns the ratings and who controls the framework.
Questions before you buy
- Can you see how a peer government rated the tool and what contract it signed?
- Are your own evaluation criteria written down, and do they include data protection and actual use?
- If an evaluation platform is run with a vendor, who owns the ratings and the criteria?
- Will your organization share its evaluations with peers?
- Who rechecks a rating when the vendor changes the product?
A shared rating is only as useful as the criteria behind it. New York’s counties have built the network. Publishing the criteria would make it a standard.
Sources