Ocean City AI Policy Sets Rules on Town Data, Tools and Wearables

A beach town writes its first rulebook for AI

Ocean City, Maryland, has adopted its first formal policy on artificial intelligence. The policy covers how town employees may use the technology, how sensitive data must be handled, which tools are approved, and how wearable devices fit in.11 According to the coverage, officials pointed to three risks when they adopted it: sensitive information getting out, AI tools inventing citations, and AI use creating public records.11

On the surface this is a small municipal story. It also captures where AI in local government stands in 2026. The technology is already in daily use, legal liability has stopped being a theoretical worry, and many small governments are writing rules after their staff and vendors have already adopted the tools.

The available reporting on the policy is thin. It confirms the main categories the policy covers and the concerns officials raised, but it does not lay out the full text or describe any enforcement mechanism.1112 Any reading of how the rules will work in practice therefore relies on context more than on the document.

The wearables clause and the police department's head start

The most telling item in the policy may be its section on wearable devices, because Ocean City's government already uses AI-capable wearables. In December, the town council approved a $4.9 million contract with Axon Enterprise to update the Ocean City Police Department's technology. The deal included AI that can analyze audio from body-worn cameras, write reports and translate.29 Officers got the new cameras over the spring. Besides translating, the cameras can answer questions about department policies and city ordinances, provide general information about the town, and transcribe audio to help draft police reports.29

A police lieutenant described officers in the field asking their cameras what an ordinance says when they can't remember it.29 That shows both the appeal and the risk. A tool that summarizes legal rules for an officer mid-encounter is a legal-tech deployment, whether or not anyone calls it one. If it gets an ordinance wrong, the town carries the consequences.

The records problem is just as real. A department communications manager said the town's records retention policy was written before body-worn cameras existed and is still under review by the Maryland State Archives. In the meantime, the department is keeping all of its footage.29 This is my interpretation and not something the reporting states, but that gap probably helps explain why public records ranked so high among officials' concerns when the council adopted the broader policy weeks later.11

Fabricated citations: a Maryland problem with a track record

Ocean City's concern about made-up citations has a clear local history. Maryland courts have dealt with AI-generated legal errors more than once. The state's intermediate appellate court referred a Bel Air divorce lawyer to the Attorney Grievance Commission after his clerk used ChatGPT to write a brief, and he admitted he had not checked the citations himself.26 That opinion was the first time either of Maryland's appellate courts addressed improper AI use by attorneys.26 In federal court, two rulings involving the same lawyer focused on filings full of nonexistent authorities. In one, the court declined to impose sanctions because the firm had promised to fix its practices.23

The state's response has tightened. In March 2026, Maryland's Standing Committee on Rules of Practice and Procedure voted unanimously to add language saying that an attorney's signature means every cited authority has been verified. The change still needs approval from the Maryland Supreme Court.26 The Maryland State Bar Association had already issued guidance in 2025 stressing human fact-checking, client confidentiality and disclosure of AI use to clients.26

The coverage diverges on one point that matters for Ocean City. A University of Maryland law professor quoted in earlier Capital Gazette reporting argued that setting AI policy for lawyers is a job for bar associations and legal institutions, not city governments.30 Read narrowly, that is correct: a town cannot discipline attorneys. Ocean City's policy covers a different group, though. Its town staff draft documents, summarize rules and answer residents' questions, and none of them are bound by professional conduct rules. In my view, the town is right to treat fabricated citations as a workplace risk and not only a courtroom one. Any employee who drafts an ordinance summary or a memo with a chatbot can introduce the same errors.

Ownership, training rights and the quieter IP question

The reporting does not say whether Ocean City's policy addresses copyright or ownership of AI output, so this is the area that calls for the most caution. The wider municipal field shows these questions are becoming standard. Among the model policies collected by Washington's Municipal Research and Services Center, one city policy covers attribution, accountability and ownership of AI outputs, and Cowlitz County's 2026 policy explicitly addresses copyright concerns.172 Mississippi's local-government template advises jurisdictions to review vendor contracts for data ownership and model-training rights before buying anything. It specifically warns against giving vendors broad rights to keep and reuse public data, and against AI features slipped quietly into existing software.33

For a town using approved-tool lists, my reading is that intellectual property will matter less through debates over who owns a chatbot's paragraph and more through procurement: whose servers hold the data, and whether prompts are used to train a vendor's models. Monroe, Wisconsin made the same calculation in October. It named Microsoft Copilot its only official tool, partly because, as reported, Copilot data stays inside the organization's own environment and is not used to train other models.15

Ocean City is one of a crowd

Ocean City is part of a broad wave of municipal AI policies, and they vary a great deal in ambition:

• Chula Vista, California voted unanimously to require public disclosure of city AI systems and human review of AI-assisted decisions. It also banned predictive policing and real-time biometric surveillance unless the council authorizes them.7
• Monroe, Wisconsin limited its policy to conversational chatbot use, and employees must get permission from the technology director to use any other tool.15
• North Ogden, Utah allows routine drafting and summarizing but requires human review and extra approval for higher-risk uses.10
• Detroit now requires a person to check AI-produced work.4
• Rapid City, South Dakota postponed its policy after a local consultant said key terms such as "sensitive data" and "appropriate review" were not defined.5

Rapid City's experience is a useful test for Ocean City. A policy that mentions sensitive data and approved tools without defining them can satisfy a council vote while leaving employees unsure what is actually allowed. Nothing published so far shows how precisely Ocean City drew those lines.

Public records: the risk most likely to bite

Of the three concerns officials named, the public records one is probably the least understood and the most likely to create legal exposure. Guidance from other states points the same way. Washington's municipal research center says AI prompts and outputs kept on agency systems are likely public records.2 A University of North Carolina analysis argues that both prompts and outputs are covered, even when they sit on a vendor's servers, as long as the government can retrieve them. The same analysis describes a Washington city employee who asked ChatGPT to write contract requirements that would quietly exclude one vendor and favor another. Those chat logs later came out through a records request.

Maryland's state-level policy already says data stored and processed in state AI systems is subject to the Maryland Public Information Act.22 That policy applies to executive-branch agencies, not to towns, but it shows how the state's legal culture views the issue.22 Because of this, Ocean City's approved-tools list does more than protect data. Requiring work accounts on sanctioned platforms is a practical way to make sure records can be retrieved when someone asks for them.

Why this matters beyond the boardwalk

Two larger pressures make policies like Ocean City's more important. The first is shadow AI. A Thomson Reuters Institute report on government legal departments found that nearly 80% of respondents worry about exposing confidential data and 73% worry about accuracy. Separately, 27% of government professionals in one survey said they had used outside AI tools at work.20 An approved-tool list is a direct answer to that behavior.

The second is liability. Commentary on the Trump administration's AI framework warns that limiting developers' liability could shift legal risk down the chain onto organizations that deploy AI models.19 Ocean City is such an organization. Meanwhile, Maryland's governor issued AI principles in September focused on human judgment and worker protections.24 So the town is setting its own rules while federal policy may leave deployers more exposed and state policy is still developing.

My reading is that Ocean City's policy is less about holding back AI than about catching up to it. The town's police already carry AI on their uniforms, its retention rules predate the devices, and its home state has been dealing with fabricated citations in court for more than a year.2926 Whether the policy works will depend on details the coverage has not yet reported: definitions, training, records capture and vendor terms.