Bluffton Audits Its Own Searches. It Cannot Audit the Other 4,000 Agencies.

Bluffton, South Carolina runs quarterly civilian audits of its license plate reader searches. It uses software that flags irregular search activity to supervisors. It has committed to a public transparency portal. Its police chief appears before council and in the press to defend the program by name. By the standards of the roughly 10,000 American jurisdictions now running these systems, that is a well governed program. It is also governing a sliver of what it actually joined.

Roughly fifty plate readers have gone up along Bluffton’s busy corridors since 2024. The department has cited fifty-two incidents the system helped resolve, thirty-eight of them cleared by arrest. Residents have pushed back anyway, opposing a proposed $170,000 camera purchase carrying $74,000 to $76,000 in annual maintenance, on privacy and racial impact grounds. Six cameras along a two mile stretch of Buckwalter Parkway were found damaged or unplugged in July. The South Carolina Attorney General has written to the vendor seeking answers.

None of that is the interesting part. The interesting part is what a civilian audit panel in Bluffton can actually see.

What local oversight reaches

A civilian audit reviews searches made by local accounts. An anomaly detector flags unusual patterns from local accounts. A transparency portal reports local totals. Every one of those controls sits on the same side of a boundary, and the boundary is the edge of town.

These networks are not local. That is the entire product. A department buys cameras, joins a national sharing arrangement, and in exchange for letting thousands of other agencies query its cameras, gets to query theirs.

Audit logs show what that means in practice. A county sheriff’s office in North Carolina released five months of network audit records this year under a public records request. Analysis of those files shows 6,389,309 searches reaching that single rural county’s cameras, run by 4,129 separate agencies. The FBI accounts for 92,716 of them. The US Postal Inspection Service accounts for 17,206. The heaviest single searcher was the Houston Police Department in Texas, more than 1,200 miles from the county it was querying, with 327,222 searches.

The structural detail matters more than the totals. In a sample of 300,000 of those searches, 84.4 percent swept more than one thousand separate camera networks simultaneously. Not one searched a single network alone.

A detective entering a plate is not querying a town. They are querying several thousand jurisdictions at once, automatically, without choosing them and frequently without knowing which ones answered.

The audit that cannot audit most of it

Put those two facts together and the limits of local oversight become concrete.

When Bluffton’s civilian panel reviews the quarterly audit, it is reviewing what Bluffton officers searched. It is not reviewing the several thousand agencies that searched Bluffton’s cameras during the same quarter. Those searches were run by people the panel cannot name, for reasons the panel cannot evaluate, under policies the panel never approved.

That is not a criticism of the panel. It is a description of what the panel was given access to. A town can audit its own conduct rigorously and still have no visibility into the majority of activity touching its own infrastructure.

The same applies to the anomaly detection software. It watches local accounts for irregular behavior. An outside agency running ten thousand queries against Bluffton cameras is not irregular behavior on a local account. It is the system working as designed.

Where the controls actually stop

Every control a department builds assumes the account is the unit of trust. Background checks vet the person who gets an account. Policies govern password sharing. Monitoring watches for strange queries from an account. Audits review what accounts did.

Recent cases in Georgia show how far that assumption travels. In August, Savannah dismissed six employees over use of the same system, including a case in which an officer’s husband, a federal immigration agent, used her account for close to a thousand plate searches across six states. In Milton, a detective was fired after running his wife’s plate seven times, a case covered here in detail. Police employees in Savannah, Milton, Augusta and Coweta County have variously been fired, arrested and sued over how the system was used.

In each case the account controls worked exactly as designed and still produced a result nobody authorized, because a credential belongs to a person, and people have spouses, emergencies, grudges and bad judgment.

Questions a council can answer before a vote

Which outside agencies can query this network today. Not the policy. The number, from the sharing list.

How many searches were run against our cameras by outside agencies in the last twelve months, and who reviewed them.

Does the civilian audit cover inbound searches from other jurisdictions, or only searches made by our own officers.

What happens to a credential when an employee leaves, and how quickly.

And the question that sounds procedural and is not: who reviews public records releases about this system before they go out, and what have they been instructed to look for.

Bluffton can answer the first four better than most towns. Very few departments anywhere can answer the last one at all.

Bluffton has 36 plate readers currently documented in the DeFlock archive with 3 more recorded as removed, against roughly fifty cited by the department. Beaufort County holds 98 active and South Carolina 2,149. Every one is listed with coordinates on the interactive map. For what these systems collect and what rights residents have, see the learn section, and for local steps see take action.

Camera counts from the DeFlock archive, community mapped and a floor rather than a census, which is why the department’s own figure is higher. Audit log analysis from network audit records released by the Surry County, North Carolina Sheriff’s Office under a public records request.