Lawmakers across the country have moved to bar local police from working with federal immigration agents. A new analysis argues that many of those bans are written so narrowly that ICE walks around them.
Immigration · State Policy
Breaking ICE: Why Most State Bans on 287(g) Agreements Leave Loopholes, and How to Close Them
Lawmakers across the country have moved to bar local police from working with federal immigration agents. A new analysis argues that many of those bans are written so narrowly that ICE walks around them.
Ban the name, and the practice survives.
That is the central warning in “Breaking ICE,” a policy analysis published October 15 by the Prison Policy Initiative and written by its legal director, Aleks Kajstura. The piece looks at state laws and executive orders meant to stop local governments from helping with immigration enforcement, and finds a consistent weakness. Many ban a specific program by its name, 287(g), and leave open the quieter ways a sheriff, a city police department or a county jail can still end up doing ICE’s work.
What 287(g) Actually Is
Section 287(g) of federal immigration law lets state and local law enforcement agencies sign agreements with Immigration and Customs Enforcement. Officers can be trained, certified or deputized to carry out parts of federal immigration enforcement, from screening people in jail to making arrests on the street. The agreements have spread quickly. According to figures cited in the analysis, agencies with active agreements now cover roughly 77.2 million people, about 32 percent of the country.
Supporters of the program, including many sheriffs and state officials, say local cooperation helps remove people accused of crimes and keeps federal resources focused where they are needed. Opponents say it turns local police into immigration agents, erodes trust in immigrant communities and drags counties into legal and financial risk. States have responded in different ways. Maine, Illinois and Massachusetts have adopted broad restrictions. Maryland has legislation barring collaboration, and the governors of Virginia and Massachusetts have issued executive orders. New Jersey passed bills of its own.
Write a ban around a program’s name, and ICE just renames the program.
Where the Bans Fall Short
The analysis identifies a set of recurring flaws. A ban that names only 287(g) leaves untouched the informal arrangements, jail contracts and information sharing that accomplish the same thing. A definition of “law enforcement” that covers only police departments lets ICE move down the list until it finds a willing partner, such as a campus force, a prosecutor’s office or a constable. Existing agreements are sometimes grandfathered, as Massachusetts did for its prisons, which keeps the collaboration alive. And carveouts based on charges can swallow the rule. New Jersey’s information sharing ban, for example, exempts law enforcement when someone has been charged with certain offenses, even without a conviction.
Then there is the matter of jails. Compliance with ICE detainers, which are requests to hold a person for the agency, and the rental of jail beds for immigration detainees keep local facilities in the custody chain even when no one has signed a 287(g) agreement. ICE also routes requests through multiple agencies at once, which can defeat a ban that restricts just one of them.
Eight Ways to Build a Stronger Ban
- Define law enforcement broadly.Cover state, county and municipal police, sheriffs, university and campus police, prosecutors and other criminal legal agencies.
- Ban the activities, not the agreements.Maine’s model prohibits a wide range of immigration enforcement activity and the use of public funds for it.
- Cancel existing agreements.Nullify them immediately or set a tight termination deadline, with no grandfathering.
- Protect sensitive locations.Restrict enforcement at schools, courthouses and hospitals, using the state’s authority over public spaces.
- Bar immigration detention in local jails.Stop state and local facilities from housing people held for immigration purposes.
- Stop honoring detainers.Prohibit local compliance with ICE requests to hold people.
- Close the federal stand-in loophole.Keep local jails out of the chain of custody when U.S. Marshals hold people referred for immigration related prosecution.
- Use every lever of executive action.Attorneys general can issue guidance on legal risks, and governors can direct state controlled agencies and create accountability commissions, as Illinois did.
What the Arrest Data Shows
The analysis leans on ICE arrest data to argue that half measures do not work. Average daily detainer related arrests grew 144 percent between 2024 and 2026. States that blocked local detainer compliance saw fewer ICE arrests overall. New Jersey, which took a compromise approach, had arrest rates comparable to Georgia, a state that enthusiastically collaborates, which the author says undercuts the idea that partial cooperation deters sweeps.
Average daily detainer related ICE arrests, 2024 compared with January through August 2026, from a companion analysis by the same organization.
Oregon offers a glimpse of how the custody loophole works. In 2026, 51 detainer related arrests there ended with a transfer to U.S. Marshals, 23 to a federal prison and 16 to local or regional jails. South Carolina shows the opposite pressure. After its attorney general encouraged participation, 287(g) agreements in the state climbed from 3 to 15 in a few months, a reminder that executive leadership cuts both ways.
A Caution on the Source
The Prison Policy Initiative is an advocacy research organization that favors restricting cooperation with ICE, and its recommendations reflect that view. The arrest numbers it uses come from ICE data, but the policy conclusions are its own. Officials who defend 287(g) will point to public safety benefits, and readers should weigh both sides when judging what a state should do.
Legislation or Executive Action?
The author’s bottom line is practical: pursue both at once. A governor’s order can move fast but reaches only state controlled agencies, and the next governor can reverse it. A law lasts longer and can reach counties and cities, but it takes a legislative session to pass and can be written with exactly the loopholes the analysis describes. Attorneys general sit between the two, able to warn local officials about legal exposure without waiting for a vote.
For voters and local officials, the useful question isn’t whether a state has a ban. It’s what the ban actually covers. Does it define law enforcement broadly? Does it cancel existing agreements? Does it keep jails out of the pipeline? A law that can’t answer yes to those questions may sound strong at a press conference and do very little on the ground.
