Curious about how the city uses AI? So were we.
Given all the debate around the safety, ethics, efficiency and pitfalls of AI, we wanted to dig into the city’s policy on artificial intelligence technology. We wanted to know all about it, like: how long Detroit has had it, what went into making it, what it covers and if it addresses concerns we’ve been hearing from residents about their data and privacy.
So, we reviewed the policy and spoke with Art Thompson, chief information officer with the city’s Department of Innovation and Technology.
Here’s what we learned.
What is AI and how is it being used by the city?
The policy is designed to promote AI use that can make city services better, while limiting risks posed to residents. The policy defines “artificial intelligence” or “AI” as a range of technologies that can make predictions, recommendations, or decisions given objectives provided by a person directly in programming.
AI systems take machine- and human-based inputs (commonly called prompts), run them through models using computer algorithms and large sets of human-generated data with the goal of creating an output like those produced by human reasoning. It is important to note that because it relies on such large amounts of human input and data, it has a tendency to replicate patterns of stereotype and bias. It can even “hallucinate” information.
Currently, the only text-based model like this (otherwise known as a Large Language Model or LLM) city employees can use is Microsoft CoPilot Chat, according to an internal memo from last month. No other generative AI or LLM — like ChatGPT, Claude, Gemini, Alexa or Siri — has been approved.
Government employees in Detroit cannot use AI to create images or media of real events, nor can it be used in evaluations of people or programs, like in employee reviews. Generally, the policy prohibits the use of AI and AI systems that handle sensitive or personally identifying information without human review.
“Detroit’s data is Detroit’s data,” Thompson told BridgeDetroit. “We own it. We do not farm it out, share it, give it out. It is a very sacred thing.”
Employees, consultants and volunteers for the city shouldn’t share data with AI that isn’t safe for public disclosure. They’re required to be clear about when they use AI, and to fact check AI outputs before using them.
You can find more specific use guidelines aligning with city values on pages 4-5. These are the city values as outlined in the policy:
1. Human-Centered Design: AI systems are developed and deployed with a human-centered approach that evaluates AI powered services for their impact on the public.
2. Security & Safety: AI systems should maintain confidentiality, integrity, and availability through safeguards that prevent unauthorized access and use. Implementation of AI systems must be reliable and safe, and minimizes risks to individuals, society, and the environment.
3. Privacy: Privacy should be preserved in all AI systems by safeguarding sensitive data from unauthorized access, disclosure, and manipulation.
4. Transparency: AI systems should value proactive communication, public disclosure and work to ensure that its data sources, operational model, and policies that govern its use are understandable and documented.
5. Equity: AI systems should effectively manage potential bias in model design, creation, and implementation and should strive to reduce harm for anyone impacted by its use.
6. Accountability: Roles and responsibilities must govern the deployment and maintenance of AI systems, and human oversight ensures adherence to relevant laws and regulations.
7. Effectiveness: AI systems should be reliable, and meet their objectives, to deliver precise and dependable outcomes for the utility and contexts in which they are deployed.
8. Workforce Empowerment: Workforce members and workforce Partners should be empowered to use AI in their roles through education, training, and collaborations that promote participation and opportunity.
9. Compliance with Applicable Laws: The use of AI systems must always comply with all federal, state, and local laws including, but not limited to the City’s Community Input Over Government Surveillance ordinance (“CIOGS”).
Will what city employees use change?
Maybe. Detroit’s policy is “flexible” and was made understanding that the technology is evolving.
Still, city employees can submit review requests for new AI technologies or any AI-inclusive upgrades to existing city systems. Requests are reviewed by several entities in DoIT, including Thompson.
Detroit has already seen the citywide coming and going of an AI system. Emily was a Duggan-era AI pilot program that answered resident calls and took messages for the Department of Neighborhoods that has since ended. Both Thompson and DON Director Bryan Peckinpaugh have said the system performed well.
How is the policy enforced?
All city departments and employees fall under the policy, which also applies to third party vendors, interns, consultants, volunteers, and more. It also extends to all AI systems and services — even the ones that don’t exist yet.
The policy is enforced by the city’s employee handbook. General violations can result in discipline of up to and including discharge; illegal activity is subject to legal action.
What went into drafting the policy?
The drafting process began in April 2023, and Thompson said it included one of the largest collaborations on a policy he’s seen. Departments like the Law Department, DoIT, Human Resources, the City Council’s Legislative Policy Division, Procurement, as well as both the police and fire departments weighed in. Thompson said a draft policy was in effect near the end of former mayor Mike Duggan’s term, but Duggan withheld final approval to allow the incoming administration to assume the lead on that front. It officially went into effect over the summer, he said.
Detroit also joined the GovAI Coalition, an international coalition of over 3,000 members and 900 government bodies, which helped inform the policy. The group promotes “responsible and purposeful AI in the public sector.”
AI has flaws. Does the policy account for them?
Yes. Aside from the risks and potential harms outlined in the guidelines, there is a dedicated “AI Risks” section of the policy. This section specifically talks about data breach risks and “adverse outcomes,” linked to AI providing bias and false information.
The requirement of human involvement in refining AI output and cautious handling of data are largely what’s meant to solve those issues.
Is Thompson concerned about anything?
Thompson’s biggest fear is about the education — or lack thereof — residents and employees do and don’t have about AI.
Thompson said he feels the technology is evolving faster than everyday users can keep up with, especially older people, children, and those who may not have access to education and tech literacy resources.
“I think it’s evolving faster than I can keep up with, and part of my job is to know what’s happening,” Thompson told BridgeDetroit. “(People are) not going to know how to protect themselves from deepfakes and a lot of the cyber security challenges that we see advancing with AI.”
Thompson said Detroit’s AI literacy ecosystem is good, but has room for improvement. He pointed to events hosted or sponsored by New Lab, Black Tech Saturdays, Rocket and more.
Though he has not seen it offer AI-specific resources, Thompson highlighted the tech literacy programming with Connect313, which also has established technology hubs in neighborhoods across the city. These hubs provide residents access to and information about technology.
I’m a resident and I have more concerns or feedback about this policy. How can I make sure I’m heard?
Input from City Council meetings are vital to how Thompson and his team understand how the public feels about the AI policy and related issues. This includes questions and concerns raised by both council members and residents during public comment periods.
So, in addition to participating in-person or virtually during formal council meetings, residents should make sure their council members know how they’re feeling. You can do this through a variety of ways, including reaching out to their offices or speaking up during district meetings that they host.
Thompson also said the Department of Neighborhoods has a “direct line,” to him and his office. You can reach out to DON district managers according to your district, during events that those managers host or attend, or during DON office/satellite hours in local buildings.