The surveillance networks cities do not have to buy

Private camera networks are beginning to emerge in cities across the U.S. which operate outside the boundaries of companies who provide surveillance cameras to local governments under contract. Rather than wait for a municipal contract, these companies install cameras through agreements with private property owners and financing from groups outside government. Police can later seek access to footage and analytical tools from systems assembled without a city purchase or public vote, but by then, the surveillance network may already be operating at scale. The emerging model does not automatically evade local law. Police access can still trigger surveillance ordinances and other restrictions. It does, however, move the initial decision to install cameras outside the municipal contracting process that has allowed communities to challenge companies such as Flock Safety . Orchestra is the clearest new example. The San Francisco startup is installing street-facing cameras free on private businesses while building an AI system capable of searching footage and following people across different locations without using facial recognition. Its expansion comes as the Los Angeles Police Department (LAPD) has raised unresolved questions about privacy, data ownership, and outside access to information collected through Flock. LAPD allowed its three-year agreement with Flock to expire July 11 and suspended its use of the company’s automated license plate readers , adding one of the nation’s largest police departments to a widening backlash against the privately built surveillance network. The contrast between government contracts with companies like Flock and surveillance cameras under private contract to businesses points to a larger problem. A city can reconsider a surveillance contract it signed, but it has far less control over a privately built network that already covers its streets. Some surveillance networks now begin with businesses that agree to host cameras. Others are financed by homeowner associations or private police foundations. The government may eventually use the resulting information, but it does not always initiate the deployment or own the infrastructure. Civil liberties groups have documented how donated equipment and free trials can place surveillance systems into operation before the reviews that accompany a conventional public purchase. This is not necessarily an unlawful attempt to evade procurement rules. Those rules generally attach when a public agency seeks money, acquires technology, or enters an agreement. A privately financed network can be built before any of those events occurs. Public review may therefore begin only when police seek access to a system that is already operating. Orchestra has deployed more than 100 cameras and plans to install 900 more along commercial corridors. The cameras stream high-definition video while AI converts what they record into searchable information. The company has said that it eventually intends to expand into every major U.S. city and then internationally. Orchestra says it does not use facial recognition. It says faces are blurred and its cameras are not placed in residential neighborhoods. The system can still link sightings of the same anonymous person by using visible details such as clothing and shoes. The technology is therefore capable of reconstructing movement without attaching a legal name to the person being followed. The company is not currently working with the San Francisco Police Department (SFPD), although it has said it is taking the steps needed to sell services to the agency. San Francisco law does not leave later police use entirely unregulated, but the rules are more complicated than a simple requirement for advance approval. Chapter 19B generally requires a city department to obtain approval for a surveillance policy before acquiring technology or entering an arrangement under which a private entity regularly supplies surveillance information. Known as the Acquisition of Surveillance Technology Ordinance, it requires all city departments to publicly disclose and seek approval for the use of surveillance technologies. A 2024 amendment allows SFPD to begin using a technology if it submits a policy within one year. Another provision says the chapter generally does not prevent city departments from receiving or using information collected by a non-city entity, except where facial recognition restrictions or an approved policy apply. A formal partnership between SFPD and Orchestra would therefore raise Chapter 19B questions. An occasional request for privately recorded footage may be treated differently. The law governs city conduct rather than serving as advance municipal approval of every private camera that could later provide evidence. Flock already operates across that public-private boundary. Its license plate readers are available to private neighborhoods and businesses as well as police departments. Flock says private customers own their information and decide whether police may access it. Its evidence policy states that privately purchased cameras can be shared with law enforcement. Ending a municipal Flock agreement therefore does not by itself remove privately owned cameras or prevent their owners from sharing records. Los Angeles demonstrated the distinction when the Flock cameras continued operating after LAPD’s regular access ended. Axon’s Fusus platform approaches the issue from the point of access rather than camera ownership. The company says its software can combine government video with feeds from commercial buildings in a single interface. A police agency can consequently obtain a much wider field of view than it could through publicly owned cameras alone. Atlanta showed how private financing can further blur the line. The Atlanta Police Foundation negotiated a multimillion-dollar purchase of Fusus technology for the police department. A later city ethics investigation found that an Atlanta police official who had previously worked for the foundation consulted for Fusus and held an interest in the company while employed by the department. The ethics office found disclosure violations and an appearance of impropriety. The official appealed. The episode did not establish that the foundation’s purchase was itself illegal. It showed how a surveillance system acquired outside a direct city-vendor contract can make responsibility and public scrutiny more difficult to trace. New Orleans provides the most consequential example of what can happen after a private network becomes intertwined with police operations. Project NOLA assembled a camera network with equipment hosted by businesses and supported through private financing. More than 200 cameras had been configured to scan faces against a watchlist. Possible matches produced automated alerts through an app used by police officers. Project NOLA had no formal contract with the city. The arrangement appeared inconsistent with a 2022 ordinance that allowed police facial recognition only for limited searches involving identified suspects in violent-crime investigations. Police Superintendent Anne Kirkpatrick ordered the automated app alerts stopped in April 2025 while the department reviewed whether they complied with the law. That action did not necessarily end the flow of information. Project NOLA staff continued receiving alerts and could pass the locations of wanted people to police through calls or electronic messages. In April, the American Civil Liberties Union said emails obtained through public-records requests indicated that officers continued sending images to Project NOLA after the announced pause. These arrangements also complicate constitutional analysis. The Fourth Amendment ordinarily restrains government conduct rather than a private party acting independently. The Supreme Court has also held that a private search can become government action when the private actor serves as an instrument or agent of the state. The determination depends heavily on the government’s participation. The high court added an important qualification in June. In Chatrie v. United States , it held that police conducted a Fourth Amendment search when they obtained a person’s Google location history data because people have a reasonable expectation of privacy in cell phone location information. The decision concerned digital location records rather than street-camera footage, so it does not decide the legality of networks such as Orchestra or Project NOLA. It does, however, undermine any broad assumption that information remains outside Fourth Amendment protection merely because a private company collected or stored it. A court examining a camera network would still have to consider what the system recorded and how completely it reconstructed a person’s movements. The government’s role in acquiring or directing the surveillance would also matter. The private model is unlikely to eliminate the backlash facing Flock. It may postpone the point at which the public can intervene. A direct city purchase creates a visible contract and often produces a public vote. A private network can grow one property at a time before government access becomes a public question. The central oversight problem is therefore no longer limited to who owns the cameras. It extends to those who may search the network and how its output becomes government evidence. Orchestra is important because it makes that transition unusually clear. The larger issue is the growth of surveillance infrastructure that a city may not buy but may eventually come to rely on.

Private camera networks are beginning to emerge in cities across the U.S. which operate outside the boundaries of companies who provide surveillance cameras to local governments under contract. Rather than wait for a municipal contract, these companies install cameras through agreements with private property owners and financing from groups outside government. Police can later seek access to footage and analytical tools from systems assembled without a city purchase or public vote, but by then, the surveillance network may already be operating at scale.

The emerging model does not automatically evade local law. Police access can still trigger surveillance ordinances and other restrictions. It does, however, move the initial decision to install cameras outside the municipal contracting process that has allowed communities to challenge companies such as Flock Safety.

Orchestra is the clearest new example. The San Francisco startup is installing street-facing cameras free on private businesses while building an AI system capable of searching footage and following people across different locations without using facial recognition. Its expansion comes as the Los Angeles Police Department (LAPD) has raised unresolved questions about privacy, data ownership, and outside access to information collected through Flock.

LAPD allowed its three-year agreement with Flock to expire July 11 and suspended its use of the company’s automated license plate readers, adding one of the nation’s largest police departments to a widening backlash against the privately built surveillance network. The contrast between government contracts with companies like Flock and surveillance cameras under private contract to businesses points to a larger problem. A city can reconsider a surveillance contract it signed, but it has far less control over a privately built network that already covers its streets.

Some surveillance networks now begin with businesses that agree to host cameras. Others are financed by homeowner associations or private police foundations. The government may eventually use the resulting information, but it does not always initiate the deployment or own the infrastructure.

Civil liberties groups have documented how donated equipment and free trials can place surveillance systems into operation before the reviews that accompany a conventional public purchase. This is not necessarily an unlawful attempt to evade procurement rules. Those rules generally attach when a public agency seeks money, acquires technology, or enters an agreement.

A privately financed network can be built before any of those events occurs. Public review may therefore begin only when police seek access to a system that is already operating. Orchestra has deployed more than 100 cameras and plans to install 900 more along commercial corridors.

The cameras stream high-definition video while AI converts what they record into searchable information. The company has said that it eventually intends to expand into every major U.S. city and then internationally. Orchestra says it does not use facial recognition.

It says faces are blurred and its cameras are not placed in residential neighborhoods. The system can still link sightings of the same anonymous person by using visible details such as clothing and shoes. The technology is therefore capable of reconstructing movement without attaching a legal name to the person being followed.

The company is not currently working with the San Francisco Police Department (SFPD), although it has said it is taking the steps needed to sell services to the agency. San Francisco law does not leave later police use entirely unregulated, but the rules are more complicated than a simple requirement for advance approval. Chapter 19B generally requires a city department to obtain approval for a surveillance policy before acquiring technology or entering an arrangement under which a private entity regularly supplies surveillance information.

Known as the Acquisition of Surveillance Technology Ordinance, it requires all city departments to publicly disclose and seek approval for the use of surveillance technologies. A 2024 amendment allows SFPD to begin using a technology if it submits a policy within one year. Another provision says the chapter generally does not prevent city departments from receiving or using information collected by a non-city entity, except where facial recognition restrictions or an approved policy apply.

A formal partnership between SFPD and Orchestra would therefore raise Chapter 19B questions. An occasional request for privately recorded footage may be treated differently. The law governs city conduct rather than serving as advance municipal approval of every private camera that could later provide evidence.

Flock already operates across that public-private boundary. Its license plate readers are available to private neighborhoods and businesses as well as police departments. Flock says private customers own their information and decide whether police may access it.

Its evidence policy states that privately purchased cameras can be shared with law enforcement. Ending a municipal Flock agreement therefore does not by itself remove privately owned cameras or prevent their owners from sharing records. Los Angeles demonstrated the distinction when the Flock cameras continued operating after LAPD’s regular access ended.

Axon’s Fusus platform approaches the issue from the point of access rather than camera ownership. The company says its software can combine government video with feeds from commercial buildings in a single interface. A police agency can consequently obtain a much wider field of view than it could through publicly owned cameras alone.

Atlanta showed how private financing can further blur the line. The Atlanta Police Foundation negotiated a multimillion-dollar purchase of Fusus technology for the police department. A later city ethics investigation found that an Atlanta police official who had previously worked for the foundation consulted for Fusus and held an interest in the company while employed by the department.

The ethics office found disclosure violations and an appearance of impropriety. The official appealed. The episode did not establish that the foundation’s purchase was itself illegal.

It showed how a surveillance system acquired outside a direct city-vendor contract can make responsibility and public scrutiny more difficult to trace. New Orleans provides the most consequential example of what can happen after a private network becomes intertwined with police operations. Project NOLA assembled a camera network with equipment hosted by businesses and supported through private financing.

More than 200 cameras had been configured to scan faces against a watchlist. Possible matches produced automated alerts through an app used by police officers. Project NOLA had no formal contract with the city.

The arrangement appeared inconsistent with a 2022 ordinance that allowed police facial recognition only for limited searches involving identified suspects in violent-crime investigations. Police Superintendent Anne Kirkpatrick ordered the automated app alerts stopped in April 2025 while the department reviewed whether they complied with the law. That action did not necessarily end the flow of information.

Project NOLA staff continued receiving alerts and could pass the locations of wanted people to police through calls or electronic messages. In April, the American Civil Liberties Union said emails obtained through public-records requests indicated that officers continued sending images to Project NOLA after the announced pause. These arrangements also complicate constitutional analysis.

The Fourth Amendment ordinarily restrains government conduct rather than a private party acting independently. The Supreme Court has also held that a private search can become government action when the private actor serves as an instrument or agent of the state. The determination depends heavily on the government’s participation.

The high court added an important qualification in June. In Chatrie v. United States, it held that police conducted a Fourth Amendment search when they obtained a person’s Google location history data because people have a reasonable expectation of privacy in cell phone location information. The decision concerned digital location records rather than street-camera footage, so it does not decide the legality of networks such as Orchestra or Project NOLA.It does, however, undermine any broad assumption that information remains outside Fourth Amendment protection merely because a private company collected or stored it.

A court examining a camera network would still have to consider what the system recorded and how completely it reconstructed a person’s movements. The government’s role in acquiring or directing the surveillance would also matter. The private model is unlikely to eliminate the backlash facing Flock.

It may postpone the point at which the public can intervene. A direct city purchase creates a visible contract and often produces a public vote. A private network can grow one property at a time before government access becomes a public question.

The central oversight problem is therefore no longer limited to who owns the cameras. It extends to those who may search the network and how its output becomes government evidence. Orchestra is important because it makes that transition unusually clear.

The larger issue is the growth of surveillance infrastructure that a city may not buy but may eventually come to rely on.