Eyes on the poor: Cameras, facial recognition watch over public housing

Surveillance cameras purchased with federal crime-fighting grants are being used to punish and evict public housing residents, sometimes for minor rule violations, a Washington Post investigation found

May 16, 2023 at 6:03 a.m. EDT
Surveillance cameras capture the daily lives of residents in New Bedford, Mass., public housing. (Video: TWP)
19 min

STEUBENVILLE, OhioWhen they installed the new surveillance system, local officials promised it would help tamp down a gang war menacing this forgotten steel town. But residents of Steubenville public housing soon learned the cameras were pointed at them.

One man was filmed spitting in a hallway. A woman was recorded removing a cart from a communal laundry room. Footage in both cases was presented to a judge to help evict the residents in court.

After the cameras caught her lending her key fob to an unauthorized guest, Melanie Otis, 52, also was threatened with eviction. Otis, who has vision loss, was allowed to stay after she explained the visitor was a friend bringing her groceries.

In public housing facilities across America, local officials are installing a new generation of powerful and pervasive surveillance systems, imposing an outsize level of scrutiny on some of the nation’s poorest citizens. Housing agencies have been purchasing the tools — some equipped with facial recognition and other artificial intelligence capabilities — with no guidance or limits on their use, though the risks are poorly understood and little evidence exists that they make communities safer.

In rural Scott County, Va., cameras equipped with facial recognition scan everyone who walks past them, looking for people barred from public housing. In New Bedford, Mass., software is used to search hours of recordings to find any movement near the doorways of residents suspected of violating overnight guest rules. And in tiny Rolette, N.D., public housing officials have installed 107 cameras to watch up to 100 residents — a number of cameras per capita approaching that found in New York’s Rikers Island jail complex.

Public housing residents are subject to outsize surveillance

Number of people per security camera

Public housing

New York

City

Omaha

Milwaukee

Rolette,

N.D.

1 camera per

19 residents

1 per

10 residents

1 per

3 residents

1.1 per

resident

Wrigley Field

Los Angeles

Airport

1 per 23 travellers

1 per 38 visitors

Rikers

Island

Louvre

Museum

Red Hawk

Casino

1 per 20 visitors

1 per 5 visitors

2 per inmate

Source: Post data collection from institutions and

security vendors

ALYSSA FOWERS/THE WASHINGTON POST

Public housing residents are subject to outsize surveillance

Number of people per security camera

Public housing

New York

City

Omaha

Milwaukee

Rolette,

N.D.

1 camera per

19 residents

1 per

10 residents

1 per

3 residents

1.1 per

resident

Wrigley Field

Los Angeles

Airport

1 per 38 visitors

1 per 23 travellers

Louvre Museum

Red Hawk Casino

Rikers

Island

1 per 20 visitors

1 per 5 visitors

2 per inmate

Source: Post data collection from institutions and security vendors

ALYSSA FOWERS/THE WASHINGTON POST

Public housing residents are subject to

outsize surveillance

Number of people per security camera

Public housing

New York City

Omaha

Milwaukee

Rolette,

N.D.

1 camera per 19 residents

1 per 10 residents

1 per 3 residents

1.1 per resident

Wrigley Field

Los Angeles Airport

1 per 23 travellers

1 per 38 visitors

Louvre Museum

Red Hawk

Casino

Rikers

Island

1 per 20 visitors

1 per 5 visitors

2 per inmate

Source: Post data collection from institutions and security vendors

ALYSSA FOWERS/THE WASHINGTON POST

Public housing residents are subject to outsize surveillance

Number of people per security camera

Public housing

New York

City

Omaha

Milwaukee

Rolette,

N.D.

1 camera per

19 residents

1 per

10 residents

1 per

3 residents

1.1 per

resident

Wrigley

Field

Los

Angeles

Airport

1 per

38 visitors

1 per

23 travelers

Rikers

Island

Louvre

Museum

Red Hawk

Casino

1 per

20 visitors

1 per

5 visitors

2 per

inmate

Source: Post data collection from institutions and security vendors

ALYSSA FOWERS/THE WASHINGTON POST

The U.S. Department of Housing and Urban Development has helped facilitate the purchase of cameras through federal crime-fighting grants. Those grants are meant to keep residents safer, and housing agencies say they do. But the cameras are also being used to generate evidence to punish and evict public housing residents, sometimes for minor violations of housing rules, according to interviews with residents and legal aid attorneys, a review of court records, and interviews and correspondence with administrators at more than 60 public housing agencies that received the grants in 27 states.

No data is available on how often the cameras are used for this purpose. But the previously unreported practice highlights how efforts to make public housing safer are subjecting many of the 1.6 million Americans who live there — overwhelmingly people of color — to round-the-clock surveillance. If evicted, former tenants can face difficulties finding housing and employment for the rest of their lives.

In an email, HUD spokeswoman Christina Wilkes said the agency never intended its safety and security grants to be used to punish residents for lease violations. But she added that such usage “is not a violation of the grant terms.”

Melody McClurg, executive director of the Jefferson Metropolitan Housing Authority in Steubenville, said tenants are responsible for upholding the terms of their lease. The cameras, which are installed only in public places, are just one way of enforcing the rules, she said, adding that other factors contributed to the JMHA’s concerns about Otis, the man who spit in the hallway and the woman who removed a laundry cart.

“People choose to get evicted by their actions,” McClurg said.

As cameras have gotten smarter, their use in public housing is becoming a flash point in the national debate over facial recognition. States including Alabama, Colorado and Virginia have passed laws limiting the use of facial recognition by law enforcement, recognizing that these tools have been shown to produce false matches — particularly when scanning women and people of color.

Last month, after The Washington Post presented HUD with evidence of the growing use of sophisticated surveillance tools by local housing authorities, the agency said it would no longer permit future recipients to spend security grants on facial recognition. These tools “are not foolproof,” and their mistakes can adversely impact public housing residents, Dominique Blom, HUD general deputy assistant secretary of public and Indian housing, said in an interview.

“This sends a signal to the housing community that this is the type of technology that the department is cautioning against,” Blom said.

An arm of the police

Public housing authorities say cameras help solve crimes and make communities safer, though few of the agencies contacted by The Post could provide data showing these results. Officials in Richmond said their new cameras helped solve 10 of the 14 homicides in public housing last year.

In Steubenville, John Stasiulewicz, a former police detective who goes by the nickname “Stosh,” is in charge of security for the public housing authority. He said the agency’s cameras have become essential to local police, who rely on him to capture footage of drug busts that have led to the arrest and eviction of dealers and users.

“I am one of the arms of the police department,” Stasiulewicz, 62, said. “I give them information and they act on it.”

Steubenville police have installed about 100 surveillance cameras across the rest of the town, a city official said, while Stasiulewicz said he monitors 161 cameras in public housing. This means public housing residents — who are nearly three times more likely to be Black than other Steubenville residents, census records show — are about 25 times more likely to have their daily lives observed by government-controlled cameras.

Stasiulewicz said he routinely uses the cameras to enforce housing rules, such as to investigate tenants who may be letting in unauthorized guests.

In Steubenville and elsewhere, public housing residents interviewed by The Post said they do worry about the safety of their neighborhoods. Some said they are happy to see more cameras outside their doors. However, many also complain that the surveillance systems “don’t work,” because they see scant evidence the devices help stop or solve crimes.

And some say cameras are being used to punish residents who pose no danger.

Tania Acabou was served an eviction notice in 2021 after the housing authority in New Bedford, Mass., used cameras to investigate her over several months. A single mother of two, Acabou had relied on her ex-husband to help take care of their kids while she worked days as a bus driver and attended night school to train for a career as a lab technician. The housing authority believed her ex was living at the house without contributing rent, court records show, violating a policy that restricts overnight visitors to 21 nights a year.

Memphis’s SkyCop cameras couldn’t prevent Tyre Nichols’s beating death

The housing authority used software to place a digital marker next to Acabou’s front door and told the system to retrieve every moment when motion was detected near the marker, documents and interviews show. When her property manager suspected that Acabou’s ex was leaving through the back door, she set up a portable camera in the backyard pointed directly at that door, according to housing authority officials and a review of the surveillance video obtained and verified by The Post.

“It got to the point where it was like harassment,” Acabou, 33, said. “They really made my life hell.”

Motorola's Avigilon software helps housing officials search through hours of video for each instance of movement in a specific area. (Video: TWP)

Sam Ackah, security director for the New Bedford Housing Authority, said the agency tries to avoid evictions by working out agreements with residents to pay their rent or come into compliance with housing rules. He said the NBHA tried to work out an accommodation with Acabou, but she refused to acknowledge that her ex was living at the apartment. Acabou said she presented evidence to the agency showing he was living elsewhere.

Ackah said his team typically conducts video investigations into 10 to 15 people at any time, working to improve life for residents who follow housing rules by policing their neighbors who don’t. The cameras — which were not purchased through a grant but through the agency’s normal annual budget — help catch people living in public housing without paying rent and help identify unregistered guests, who have not been vetted for past offenses, he said.

Unlike the doorbell cameras voluntarily installed by residents of more affluent communities, surveillance cameras typically are installed in public housing without the consent or control of the residents. In D.C., one woman and her son were arrested in 2018 after trying to stop housing officials from entering her apartment to install a camera power box in her bedroom. In a lawsuit filed against the city, the woman claims a security officer who arrested her said she “did not have any rights as a public housing resident and that she could not stop the worker from installing the cameras.”

Lawsuit alleges D.C. Housing’s cameras could ‘capture intimate details’

Rachel Molly Joseph, chief operating officer for the DC Housing Authority, declined to comment on the lawsuit, which is pending, but said none of the cameras face the interior of residents’ homes.

In Charlottesville, housing officials met with residents before installing a new surveillance system and learned they didn’t like the idea of officials “just sitting there watching the cameras to see if tenants are breaking their lease,” John Sales, the agency’s executive director, said in an interview. Unlike many housing authorities contacted by The Post, officials in Charlottesville pledged not to use the cameras to enforce lease agreements.

Nationally, evictions from public housing have surged since late 2021, when the federal government lifted an eviction moratorium put in place to protect renters during the pandemic, according to Peter Hepburn, associate director of Princeton University’s Eviction Lab. In the 10 states and 34 cities his group tracks, there were at least 5,576 evictions by public housing authorities in 2022, about twice as many as the previous year. Public housing evictions grew at a faster rate in that period than evictions overall, Hepburn said.

There is no way to measure how much of the increase may be due to more sophisticated surveillance. But attorneys who defend tenants in eviction cases said that in some communities with newer cameras, they have seen an uptick in cases referencing video footage as evidence for kicking people out.

Gavin Bates, a legal aid attorney in New Bedford, said the local housing authority’s surveillance system now “regularly appears” in cases where his clients are being evicted. The authority uses this system “to great effect in moving people out,” he added.

“Quite often when there is camera footage of an event, the event does show a rules violation of some kind,” Bates said. “But there are also a lot of unrepresented folks who just believe that there is camera footage when they are told, the footage is never produced, and they do not know their rights and often make bad decisions as a result.”

Ackah said the New Bedford Housing Authority provides video evidence to the court in cases where it exists. “If we do not have the evidence to prove a lease violation, then we would not be able to evict,” he said.

Surveillance by algorithm

Camera manufacturers claim that artificial intelligence has made their devices capable of doing much more on their own, from identifying faces and license plates to spotting people loitering. As a result, residents in a growing number of cities are being watched by computer algorithms programmed to help authorities make split-second decisions about who and what is suspicious.

Of 41 housing authorities that told The Post they had bought new cameras using a HUD grant in recent years, 11 said their systems were equipped with facial recognition tools. Six said they planned to use this capability, either to assist police in criminal investigations, to control tenant access to buildings or to actively search for people banned from their properties.

Housing officials in Scott County, Va., said they use cameras to spot former tenants and guests with some previous violation involving drugs or violence. After getting a $250,000 HUD grant and installing 80 cameras — about 1 for every 2 public housing residents — officials in Scott County uploaded photos of all the people they had barred from entering public housing. (Such ban lists, which are widely used in public housing, are controversial because authorities often target people who have never been convicted of a crime.)

Since the cameras began to actively look for banned individuals, they have made only one match, according to Sue Smith, executive director for the Scott County Redevelopment & Housing Authority. It correctly identified a man banned from the property because of a past domestic violence incident involving a resident, an incident that did not result in criminal charges, Smith said. As soon as the match was made, the software sent an alert to housing staff, who asked the man to leave and alerted the county sheriff.

Trusting computer algorithms to determine who or what is suspicious is likely to put marginalized groups at greater risk of wrongful arrest, said Patrick Clark, managing counsel of the legal department at the Office of the Ohio Public Defender. A 2019 study of facial recognition systems found that Asian and African American people were up to 100 times more likely to be misidentified than White men.

“If the information getting spit out by the algorithm is not accurate or not fully accurate, but you have law enforcement or agents of the government acting on a real time basis, it puts people at risk of loss of liberty or worse,” said Clark, who was part of a state task force that recommended restrictions on the use of facial recognition by law enforcement.

FBI, Pentagon helped research facial recognition for street cameras, drones

The cameras in Steubenville and Scott County were made by Verkada, a Silicon Valley start-up whose entry-level package — starting at less than $1,200 for a single camera and a one-year software subscription — comes preloaded with facial recognition.

Earlier this year, surveillance industry researcher IPVM tested a Verkada camera’s ability to scan a group of people and identify any matches with a database of faces. About 15 percent of the matches were wrong. When the people were wearing masks or viewed at an angle, Verkada got up to 85 percent of matches wrong.

Brandon Davito, vice president of product at Verkada, said the company’s own testing found that its facial recognition system accurately matched more than 99 percent of faces. He said IPVM’s test failed to use the most precise setting of its software.

Smith said she believes that if the Scott County system falsely identifies an innocent person, her staff and the police would be able to sort it out before any harm was done.

A 2022 law prohibits police in Virginia from using facial recognition to track “the movements of an identified individual in a public space in real time.” But the law only applies to state, local and campus police, so the Scott County housing authority likely is not in violation, said state Sen. Scott A. Surovell (D-Fairfax), who wrote the statute.

Last year, Verkada disabled facial recognition capabilities in its cameras in Illinois and Texas, where companies have recently been charged with violating state laws prohibiting the collection of biometric data without user consent. Davito said that when new customers turn on a Verkada camera, they have to check a box confirming that they are using facial recognition in accordance with local laws before the feature is activated.

Texas sues Facebook parent Meta over use of facial recognition

In an April 21 public notice, HUD announced a ban on “automated surveillance and facial recognition technology” — terms that the agency did not define. The restriction applies only to future recipients of its security grants and does not limit use of surveillance tools by authorities that have already purchased them, said Blom, the HUD official. The agency is still assessing the need for a broader ban.

The action came four years after a group of Democratic lawmakers led by Sen. Ron Wyden (Ore.) urged HUD to become more proactive in setting limits on surveillance. “There is very little evidence at this point to make a judgment that these systems are making people safer,” Wyden said in an interview last month.

In response to the lawmakers, a HUD official said in 2020 that the agency would evaluate how facial recognition was being used in public housing. But Wilkes, the HUD spokeswoman, said the agency has yet to do research on the topic.

Cameras, not counseling

The HUD security grants were funded by Congress in 2009 as part of an omnibus appropriations bill. They were intended in part to offset the loss of the agency’s Drug Elimination Program, an expansive federal initiative launched in 1989 at the height of the crack cocaine epidemic. The program ended in 2002.

The Drug Elimination Program provided more than $100 million a year for a wide range of social services, such as drug rehabilitation, in addition to security improvements. The current security grants, by contrast, provide just $10 million a year primarily for equipment such as cameras, doors, lighting and smoke and carbon monoxide detectors. HUD also permits housing authorities to allocate some of their annual capital funds for these items.

This shift in federal policy and reduction in funding forced some housing authorities to stop helping residents struggling with addictions and mental health problems, said David Weber, a senior policy analyst at the Public Housing Authorities Directors Association, a Washington-based trade group.

“How do you address a drug and crime problem when you are having trouble keeping the heat on and the elevators working?” Weber asked.

In some places, public housing officials say cameras have become one of their best tools for addressing the crime and drug problems plaguing their communities.

When the Jefferson Metropolitan Housing Authority applied for its federal security grant in 2019, it said it was desperate to restore order to the chaotic lives of its tenants. Across Steubenville, a former manufacturing hub 40 miles west of Pittsburgh, methamphetamine overdoses were on the rise, two rival gangs were shooting up apartments in turf disputes, and a 15-year-old boy had been killed.

Michael Thomas, a community advocate in Steubenville, said the town’s drug and gang activity stems from a lack of investment in local schools, arts, commerce and the kinds of recreational activities that would help young people improve their lives. Because the city offers so little support for people dealing with addiction and mental health issues, many of them end up on the street, he said.

McClurg, a 49-year-old former health-care administrator who fills her office with inspirational quotes from Mother Teresa, said she would like to do more to help tenants, but she doesn’t have the funds for new services. It pains her to evict tenants, she said, but she can’t let them bring drugs into the community and pose a danger to other residents.

“Number one, we are a landlord,” McClurg said.

At the news conference announcing the $250,000 HUD grant that financed the new surveillance system, county prosecutor Jane Hanlin told the Daily Herald newspaper that it would help improve public safety in the area.

“It will work as a crime deterrent and a crime solver,” she said, “so there’s no down side.”

Some public housing residents have seen the downside of living under surveillance. Acabou, the single mother in New Bedford, Mass., said her landlord used cameras to systematically scrutinize her life and pressure her to leave.

A video clip submitted by New Bedford Housing Authority as part of its case to evict Tania Acabou. (Video: TWP)

In 2021, the housing authority gave her lawyer a thumb drive filled with hours of surveillance video — scenes of her ex-husband providing parenting and support to their two kids. She struck a deal with the housing authority to voluntarily leave public housing to avoid having a formal eviction on her record.

Her new apartment is smaller and more expensive.

For others, the consequences of increased surveillance have been more severe.

Deborah Seekins was evicted from her high-rise apartment in New Bedford in 2021, after the housing authority said its cameras repeatedly caught her smoking cigarettes too close to the building and getting into altercations with other tenants, court records show.

Seekins, 68, now sleeps in her sister’s basement. She is trying to find a place of her own and recently had an application for a new apartment accepted. But then the landlord called public housing for a reference check, and Seekins was told the offer had been rescinded.

“I don’t know what’s going to happen now,” she said.

Jonathan Baran in San Francisco, and Monika Mathur and Magda Jean-Louis in Washington contributed to this report.


About this story
The Post contacted more than 85 public housing authorities that had accepted a HUD Emergency Safety and Security Grant in fiscal years 2021 and 2022 and had indicated they planned to spend those funds on cameras. Agencies were asked to identify the make and model of the cameras they purchased, whether those tools were equipped with facial recognition and whether they planned to use facial recognition or any other video analytics tools.
Of the 60 agencies that responded, 40 confirmed that they had already purchased cameras. Of those, 10 said they had facial recognition and five said they planned to use it. One additional agency awarded a HUD grant in 2019, Jefferson Metropolitan Housing Authority, also confirmed that it purchased facial recognition tools and planned to use them.
Additionally, The Post asked some housing authorities for their total numbers of residents and cameras and divided those numbers to see how many residents are typically watched by each camera. Of the 27 agencies that shared these figures, the average ratio was 1 camera to every 14 residents.
Chart data does not take into account the size of spaces under surveillance, which usually factors into the number of cameras necessary, security experts said.