public-private partnerships are central.

carceral AI runs on public-private partnerships.

Carceral AI is part of a broader trend of the privatization of public sector systems, including the criminal legal system, child welfare, immigration, and education. The rise of neoliberal governance – a political-economic approach to market-oriented reform favoring deregulation and privatization – has led to outsourcing portions of work to private companies that then contract for the government.

In some cases, government agencies hire consultant companies to build in-house programs, such as New Orleans police hiring Palantir for predictive policing software and the NYPD hiring Microsoft to build their Domain Awareness System, a mass surveillance system. In other cases, government agencies rent or subscribe to services provided by external vendors that specialize in creating and marketing products specifically to police, courts, prisons, and immigration agencies. Some of the key vendors include Clearview AI, which provides facial recognition programs to police; Sound Thinking, which provides the audio gunshot locator Shotspotter; Geolitica, formerly known as PredPol, which provides predictive policing software; Equivant, formerly known as Northpointe, which created the notorious COMPAS risk assessment; STRmix and TrueAllele, which provide probabilistic DNA profiling software to forensic labs; Vigilant Solutions and Flock Safety, which create and rent automatic license plate readers to cities and police departments; private prison corporations GEO Group and CoreCivic provide electronic monitoring and “community-corrections” services to prison systems; and Palantir, which contracts software to Immigration and Customs Enforcement (ICE) to arrest parents of unaccompanied migrant children.

Companies present their products as more scientific, objective, and data-based than human decision-making, boasting the quantity of data at their disposal and the complexity of their algorithms. Some companies offer data sharing across police departments, allowing inter-city collaborations that were previously infeasible. The involvement of private companies also introduces a new layer of opacity: whereas tools built within government agencies are often subject to FOIA requests or public sharing agreements, proprietary tools do not have the same transparency requirements. This can lead to software that exacerbates existing inequalities, either by its design or due to undiscovered errors in its code.

In response to critique, companies routinely rebrand (Predpol becomes Geolitica, Northpointe becomes Equivant, ShotSpotter becomes SoundThinking). They also consolidate power through parent companies and expand into other surveillance technology markets, such as Flock Safety launching gunshot detection that pairs with their automatic license plate readers.

Public-private surveillance partnerships extend beyond companies that market directly to criminal legal system agencies. Flock Safety partners with homeowner associations that then share data with law enforcement as well. Amazon Ring, the video doorbell, sells homeowners’ camera footage to police departments in partnership agreements. Social media companies also share information, including in private messages, with police departments. This process of “infrastructural obfuscation” hides surveillance technology as part of the community infrastructure, blurring the boundaries between police and civilian surveillance.