AI can genuinely transform how local governments understand and care for their communities, but that promise only holds if the technology is built responsibly from the ground up. At City Detect, that's not a marketing line. It's a set of commitments we hold ourselves to from design to development, and we want to be transparent about what they are.

Our Responsible AI Strategy is our commitment to the public, especially our partners. Here's what it says and means in practice.

No tracking of personal information. Faces and license plates are blurred by default. PASS AI® is a built-environment tool. It's designed to understand property conditions and roadside conditions.

No data sharing. City Detect does NOT share any jurisdiction’s data with any third party.

No federal or law enforcement database ties. City Detect does not tie into federal or law enforcement databases.

No revenue from detections. Neighborhood services departments don't operate on quotas. City Detect never collects revenue based on how many notices go out, and we don't send them automatically through our system.

Human in the loop. Local leaders understand their communities best, and they know how to administer neighborhood services well. That's why every single detection requires human review before any action is taken. The AI surfaces what it sees. A person decides what happens next.

Stateside computation and storage. Our cloud computation and data storage infrastructure is located in the United States.

Annual SOC 2 audit. An independent SOC 2 audit verifies our security and data-handling controls every year. We achieved SOC 2 Type I compliance in 2024 and SOC 2 Type II compliance in July 2025, which we’ve maintained in 2026. This evaluation is conducted by an independent firm against the AICPA Trust Services Criteria across security, confidentiality, availability, processing integrity, and privacy.

Why This Matters More Than Ever

AI bias happens when a system produces skewed results because of flawed assumptions or unrepresentative data. In government applications, that kind of bias can shape which neighborhoods get resources and which get overlooked. We take that seriously, and it shapes how we build.

Population-level data, not samples. Many manual and AI systems train and operate on sampled data, which can bake in the very disparities we're trying to avoid. Our training data is aggregated across dozens of cities and neighborhoods spanning a wide range of environments and income levels, so our models aren't skewed toward any one type of community. In practice, coverage depends on how many detection units a jurisdiction deploys and where they travel. We work closely with every jurisdiction to ethically and equitably deploy PASS AI® so that no neighborhood is overlooked or receives disproportionate attention. 

A transparent, rigorous review process. Reviewers trained on our specific detection categories, such as litter, graffiti, or storm damage, evaluate model output using statistically rigorous sampling techniques. We don't deploy a fully automated model to production until it achieves at least an 85% F1 Score, a metric that requires both high precision and high recall. Any corrections our reviewers make feed back into training the next version of the model.

Explainable by design. We favor simple, interpretable models over opaque ones. Every detection comes with a confidence score between 0 and 1, so users can see exactly how sure the model is, not just what it concluded. Detections below a set confidence threshold never even reach the platform. That's a deliberate choice, and it's one reason PASS AI® doesn't produce the kind of "black box" outputs people rightly worry about with generative AI.

Built on real-world understanding, thoughtfully. Our training data is collected through City Detect's own camera units and reviewed by experts trained specifically on the object categories we detect. We also train our models under a range of conditions, such as unusual weather or less common objects of interest, so performance holds up across a wider range of real-world scenarios.

Ongoing Commitment, Not a One-Time Fix

City Detect believes that Responsible AI is not a box to be checked once; it's an ongoing discipline to be maintained.  That's why City Detect participates in the GovAI Coalition, a nationwide initiative of more than 1,000 members and 350-plus local, state, and federal agencies working toward more transparent, accountable AI in the public sector.

Our full Responsible AI Strategy is organized around four pillars: legality, data integrity and privacy, ethical design and development, and continuous improvement. You can read the complete strategy at citydetect.com/responsible-ai-strategy.

AI in public service isn't just about efficiency; it’s about earning and keeping the trust of the communities these tools are meant to serve. We're proud to be transparent about how we're doing that, and we’re committed to continuing to build in this way.