Thursday, November 19, 2026

U.S. Farmers gear up for war over leaked bill that could open door to land seizures

By: Daniel Hampton
Nov 18, 2026
RAW STORY

Farmers gear up for war over leaked bill that could open door to land seizures
Demonstrators chant and hold signs during a protest against data centers outside Rockford City Hall in Rockford, Illinois, US, November 17, 2026. REUTERS/Yong Li Xuan


Ohio farmers are mobilizing against a leaked draft bill they fear could let energy developers take their land for data centers, just two weeks after voters in 18 Ohio communities voted against the projects.

The draft, obtained by WEWS, would expand Ohio’s definition of a “public utility” to include developers holding a state Power Siting Board certificate for electric transmission lines or gas pipelines. According to legal experts cited by the outlet, a transmission developer could then seize land to power or host a data center. The draft would also let those companies use a fast-track process called “quick take,” WDTN reported. Under quick take, a company deposits money, takes the property, and settles the final price later in court.

The Ohio Farm Bureau has braced for the proposal since June, when the Ohio Business Roundtable’s push to change eminent domain law came to light. The group says it does not oppose data centers.

“We are aware of efforts to further erode the limited protections that landowners have, allowing for quick take of property without first paying for the property and determining a landowner’s rights and compensation through a court of law,” the Bureau’s Evan Callicoat said.

Gretchan Francis, whose family farm has already lost land to highway projects, said losing land to a data center would be devastating.

“I think anybody would feel the same way if their personal land was taken for a data center,” she said.

Champaign County farmer Josh Gentis told WDTN that lawmakers should take heed of the election, when Ohio voters went 21 for 21 against data centers. Farmers are preparing a calling campaign to pressure legislators to kill the bill.

Republicans have fretted over the issue since at least August, with a National Republican Senatorial Committee memo telling AI companies that backlash threatened the party’s hold on an Ohio Senate seat. Even former Trump voters in Ohio soured on the president over his support for the facilities, and national polling has shown broad opposition across party lines.

Incoming state Senate President Bill Reineke (R-Tiffin), who is working on the bill, said it would not extend eminent domain to anyone who lacks that power today. He called it a framework to bring the state in line with 45 other states. Alexandra Denney of the Ohio Business Roundtable said the goal is a reliable, affordable grid and that the group does not intend to expand eminent domain. After the draft was read to her, she said she would need to look at the public utility language.

The bill has not been formally introduced.
Trump’s GOP law ripped food aid from millions of kids faster than anyone feared: analysis


By: Brad Reed,
 Common Dreams
Nov 18, 2026

Trump’s GOP law ripped food aid from millions of kids faster than anyone feared: analysis
FILE PHOTO: Volunteers place food items in vehicles during a mobile food distribution at Cedar Creek High School, as nearly 42 million Americans face a potential lapse in Supplemental Nutrition Assistance Program (SNAP) benefits, known as food stamps, due to the second-longest U.S. government shutdown, in Cedar Creek, Texas, U.S., November 1, 2025. REUTERS/Kaylee Greenlee/File Photo


Cuts to federal food assistance in Republicans’ 2025 budget law are leading to drops in participation that are significantly higher than those projected by the Congressional Budget Office, according to an analysis published on Tuesday by economist Diane Schanzenbach.


Schanzenbach, a professor at the McCourt School of Public Policy at Georgetown University, found that participation in the Supplemental Nutrition Assistance Program (SNAP) fell by 13% in the year after President Donald Trump signed the One Big Beautiful Bill Act in

The decline in SNAP enrollment has been “much faster and deeper than anticipated,” Schanzenbach wrote, and participation in the program is now at its lowest rate since 2009.

“SNAP participation fell in every state except Alaska,” the economist explained. “The size of the drops varied widely: over 50% in Arizona, about 20% in Texas and 22% in Florida, compared with 7% in California and 9% in New York.”

Given that the unemployment rate has gone up since the start of Trump’s second term, there is no chance that drops of this size can be attributed to normal changes in economic conditions, Schanzenbach added, making it likely the GOP budget law’s cuts to the program are a main culprit.

The most high-profile change to the program was expanding work requirements to more recipients, but Schanzenbach found that these new eligibility rules alone “don’t account for the full decline.”

“Families with children are largely outside the new work requirements, yet an estimated 2 million fewer children received SNAP in October 2026 than in July 2025,” she explained. “This likely reflects some combination of eligible households lost at renewal over paperwork, fewer new applicants, and immigrant families staying away for fear of immigration enforcement, including members who remain eligible.”

State governments are another important variable determining SNAP participation declines, and Schanzenbach pointed to Arizona as an example of what happens when a state moves aggressively to reduce the risk of incurring financial penalties under the budget law’s payment error rate provisions.

“Arizona moved earliest and most aggressively to cut its error rate, and its caseload fell by about half between July 2025 and July 2026, the steepest drop in the country,” she wrote. “At the same time, the number of state staff determining eligibility fell from 1,370 in July 2024 to 880 a year later. Applicants described phone lines that wouldn’t connect, being informed of interview appointments after they were scheduled to occur, and cases closed over a single wrong document.”

The impact of these cuts on hunger is now more difficult to determine, Schanzenbach said, because the US Department of Agriculture no longer conducts an annual Food Security Survey.

That said, the economist projected that “losing SNAP benefits is likely to raise food insecurity for many families,” while pointing to the increased activity seen at food banks across the US as evidence of the cuts’ impact.

“In the face of these dramatic reductions in SNAP participation,” Schanzenbach wrote in her conclusion, “it becomes even more important to restore the collection of national food insecurity data so that policymakers can assess whether the shrinking caseload means more families are going without enough to eat.”
Cops swarmed teens with guns drawn after AI mistook Doritos for a 'gun': lawsuit


By:William J. Ford,
Nov 18, 2026


After an artificial intelligence program mistakenly identified a Baltimore County high school student’s bag of chips as a gun, county police and school officials turned that mistake into a “terrifying” and “life-threatening” encounter for the four teens involved, a new lawsuit says.

The suit was filed in Baltimore County Circuit Court Tuesday, a little more than a year after the confrontation at Kenwood High School. The 31-page complaint claims police officers inflicted pain and suffering on Oct. 20, 2025, on Taki Allen, then 16, and his three friends – Yui-Osa Ogbeide, Devon Thompson Jr. and Keith Winfield III – all now 18.

The incident began about 6:30 p.m., as the four were talking and waiting outside the school for a ride after football practice. Allen had a bag of Doritos that his coach had given him after practice, a bag that the school’s Omnilert Gun Detection identified as a gun, alerting school officials.

Three officials who received the alert recognized the mistake and canceled the alert within minutes. But the school’s principal, Catherine Smith, “never acknowledged the Alert’s cancelation or took any steps to verify whether the Alert remained active,” the suit says.

Instead, more than 10 minutes after the alert was canceled, she sent a screenshot of video footage of Allen “holding the chip bag” and sent a text message to the school resource officer, who forwarded the screenshot to the police department, prompting “an armed police response.”

The alert went out about 7:05 p.m. By 7:21 p.m., six police cruisers surrounded the teens and officers got out, “many with guns drawn,” according to the complaint. They forced the teens to the ground, cuffed and searched them, according to the suit, which shows a picture of the boys with their arms raised outside the school.

After they determined there was no gun, one officer told the students that “AI is not the best,” and another muttered “stupid AI cameras” and “[expletive] Omnilerts,” according to the complaint. “Yet officers relied on the Alert to initiate this painful encounter with the students,” it states.

One of the officers, Corp. Michael Foehrkolb, allegedly told the teens, “Hey, thanks for not running away, that would have been a real problem,” the suit says. It shows a text message from the principal to school staff and administrators that “made light of the incident.”

The text states: “No worries. SRO (school resource officer) had patrol car roll through. It was a bag of chips.” The message included four emojis crying from laughter.

Freya Jamison, an associate counsel with the Lawyers’ Committee’s Criminal Justice Program, said in an interview Tuesday there’s been no public apology from the school’s principal to the teenagers or their families. Jamison said Smith remains at the school.


“We can’t ignore the potential consequences of an armed police encounter, especially with young Black men like Takei, Osa, Devon, and Keith,” Jamison said. “They survived the encounter, but their safety never should have depended on luck.”

“As AI is on the rise around the country, it’s critical that technology companies and public officials are held accountable when they use that technology to violate constitutional rights,” she said. “We simply can’t outsource human judgment to machines, and this case shows exactly why.”

Omnilert spokesperson Blake Mitchell said in an email Tuesday that the company, named as a defendant, “cannot comment on lawsuits.” Baltimore County police and county school officials also declined to comment on pending litigation.

‘Failure’

The complaint notes that Omnilert’s AI weapon detection system was patented in 2022, “barely a year before it was deployed” in Baltimore County Public Schools, which was among “the first of its initial deployments.”

Richard Henry, the Maryland Office of Inspector General of Education, said in December on the “C4 and Bryan Nehman” radio show on WBAL that Omnilert works in Baltimore County schools. Henry attributed the Kenwood High situation to “human error” when looking at video and still images on a cell phone.

“In this case, this occurred in the evening hours, so it was low light,” Henry said on the show. “The student had the bag next to their hip, and they had the way that their hand was situated [with] a couple fingers in, a couple fingers out. That movement was detected, and it was just more or less, this is an alert. It could be a possible threat.”


But the complaint notes that Omnilert’s AI software suffered from false negatives before, including in January 2025 when it failed to detect a shooter’s gun at Antioch High School in Nashville. Two students died, including the shooter, and another was shot in the school’s cafeteria in that incident.

All four teens in the Kenwood High incident remain friends, the complaint says. Allen is still at the school, while Ogbeide, Thompson and Winfield, who graduated earlier this year, plan to visit the school on breaks from college and attend basketball games and other activities.

But with the current AI system in place and without adequate training for police and school staff, there remains concern for all four teenagers’ safety.


“The policies, customs, and training failures that produced this incident also remain in place, putting all four Plaintiffs at ongoing risk of being detained and searched without reasonable articulable suspicion,” the complaint states.

Maryland Matters is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Maryland Matters maintains editorial independence. Contact Editor Steve Crane for questions: info@marylandmatters.org.

Wednesday, November 18, 2026

Substance use, mental health, and respiratory infections shifting burden of disease in Canada



Analysis of Global Burden of Disease Study of 204 countries looks at Canadian trends


Canadian Medical Association Journal






The conditions driving the burden of disease in Canada are shifting, according to an analysis of the Global Burden of Disease (GBD) Study published in CMAJ (Canadian Medical Association Journal) https://www.cmaj.ca/lookup/doi/10.1503/cmaj.260996. Cancers, cardiovascular diseases, musculoskeletal disorders, and unintentional injuries were the four leading causes of health loss in Canada in 2023, accounting for half of all health loss. But the fastest growing contributors between 2010 and 2023 were substance use disorders, mental health conditions, and respiratory infections, including COVID-19, after accounting for age.

The GBD Study, led by the Institute for Health Metrics and Evaluation at the University of Washington, estimates health loss from diseases, injuries, and risk factors in 204 countries and territories. It covers 292 causes of death, 375 diseases and injuries, and 88 modifiable risk factors. A team led by Canadian researchers analyzed these estimates to describe health loss in Canada in 2023, trends from 2010 to 2023, and how Canada compares with other G7 countries.

To measure health loss, the study uses disability-adjusted life years (DALYs). One DALY represents one year of healthy life lost, whether due to premature death or to disability from illness or injury. The higher the number, the greater the health loss.

“The point of measuring burden this way is to see where health loss is growing, not just where it is largest. Cancer and cardiovascular disease remain the leading causes of health loss, but they are declining once you account for an aging population. Mental health and substance use disorders are moving the other way,” says Dr. Heather Orpana, senior research scientist at the Public Health Agency of Canada and adjunct professor at the University of Ottawa, Ottawa, Ontario. “Unlike many of the leading causes of health loss, mental health and substance use disorders tend to affect people earlier in life, during the years when they are finishing school and starting work. That means the impact goes well beyond what we measure in a health study.”

Highlights:

  • Noncommunicable diseases were responsible for the vast majority of all DALYs, making up 82% of the disease burden in Canada in 2023.
  • Cancer (neoplasms) was the leading cause of all-age DALY rates in both males and females.
  • Mental disorders were the fifth leading cause of health loss for both sexes combined.
  • If mental disorders, substance use disorders, and self-harm were combined, this group would be the second leading cause of health loss, behind cancer.
  • Health loss due to substance use disorders had the largest percentage increase (90%) between 2010 and 2023.
  • There were also large increases (67%) in health loss from respiratory infections, including COVID-19 and tuberculosis, from 2010 to 2023. COVID-19, absent before 2020, was responsible for most of the increase in the all-age DALY rate from 2010 to 2023.
  • For females, 57% of health loss came from years lived with disability; for males, 55% came from premature death. About 40% of total health loss as measured by DALYs could be attributable to risk factors that were measured in the GBD study.
  • Behavioural risk factors such as tobacco use or low physical activity accounted for about a quarter of all-age DALYs, and metabolic risk factors such as high cholesterol or high blood sugar accounted for almost 20%.
  • Environmental and occupational risks, like lead exposure or air pollution accounted for 6%.
  • Totals don’t add up due to overlapping risks and mediating pathways.
  • Tobacco use continued to be the leading risk factor contributing to all-age DALY rates in 2023. Although there was no significant change in all-age attributable DALYs due to tobacco use between 2010 and 2023, after accounting for the aging population, the attributable DALY rate due to tobacco use decreased by almost 20%.
  • All-age DALYs attributable to drug use as a risk factor increased by 118% between 2010 and 2023.

“Almost 40% of the health loss in Canada can be traced to risk factors we already work hard to address: things like tobacco, diet, blood pressure, body weight, and physical inactivity. That portion of the burden is, in principle, preventable,” says coauthor Dr. Justin Lang, research scientist at the Public Health Agency of Canada. “Many of the risk factors we measure, like diet or physical inactivity, are shaped by conditions people don’t fully control. Focusing only on individual behaviour can draw attention away from those upstream causes.”

Canada’s overall burden of disease was comparable to that of other G7 countries, while Canada’s life expectancy fell in the middle. The United States had the highest age-standardized DALY rate in G7 countries, and Japan and Italy had the lowest. Health loss due to substance use disorders was ranked substantially higher in Canada and the US, and ranks for cardiovascular disease and cancer (neoplasms) were similar among G7 countries.

 

Mutual-help group attendance in alcohol treatment drops sharply, with wide racial and ethnic gaps



National study of nearly 2.8 million treatment discharges finds the steepest drops in weekly attendance came during the COVID-19 pandemic




Public Health Institute





Oakland, CA—Among people in inpatient alcohol treatment, the share attending Alcoholics Anonymous or similar groups at least weekly fell from more than half to fewer than four in 10 in just seven years, according to a study published in the Journal of Studies on Alcohol and Drugs. Latinx/Hispanic, Black and American Indian/Alaska Native patients were less likely than White patients to attend weekly, even after accounting for factors such as age, employment and alcohol use.

Researchers analyzed federal data on nearly 2.8 million discharges from publicly funded substance use treatment facilities in every U.S. state except Oregon between 2015 and 2022. The study includes adults treated for alcohol use, either alone or alongside other drug use. Analyses examined inpatients and outpatients separately.

“Mutual-help groups are widely available, low-cost and one of the most effective resources we have for supporting long-term recovery from alcohol use disorder,” said Sarah E. Zemore, Ph.D., lead author and senior scientist at the Alcohol Research Group, a program of the Public Health Institute. “Attendance has dropped nationally, and it is considerably lower for several racial and ethnic groups. Treatment programs are well placed to change that by actively connecting patients with meetings where they feel they belong.”

A steep national decline

Weekly attendance, defined as attending four or more meetings in the 30 days before discharge, dropped substantially over the study period:

  • Inpatient treatment: from 53.9% in 2015 to 39.2% in 2022
  • Outpatient treatment: from 19.7% in 2015 to 16.0% in 2021 and 2022

After accounting for other factors, the odds of weekly attendance in 2022, relative to 2015, were about one-third lower for outpatients and about one-half lower for inpatients. The sharpest drops came in 2020, when the COVID-19 pandemic limited in-person meetings. For example, inpatient attendance fell from 49.8% in 2019 to 41.7% in 2020 and remained low through 2022.

The researchers suggest that a shift in the treatment field away from abstinence-focused goals may also have played a role, along with wider use of medications for substance use, which can be stigmatized in some 12-step programs.

“Medications and mutual-help groups aren’t competing options. Many people do best with both, so programs should encourage support group participation whether or not a person is taking medication,” said Katherine Karriker-Jaffe, Ph.D., co-author and senior director at RTI International, an independent scientific research institute.

Gaps across racial and ethnic groups

Compared with White patients, odds of weekly attendance across settings were:

  • 36% to 39% lower for Latinx/Hispanic patients
  • 28% lower for Black patients
  • 22% to 28% lower for American Indian/Alaska Native patients
  • 10% to 11% lower for patients of other races and ethnicities

In contrast, Asian/Pacific Islander patients had 60% to 66% higher odds of weekly attendance than White patients. Women had slightly higher odds of weekly attendance than men. Compared with adults ages 45-54, young adults (18–24) and adults 65 and older had lower odds of weekly attendance, as did those who were married (vs. unmarried), employed (vs. unemployed) and those without a college degree.

Why it matters

Rigorous research shows that participation in AA and similar groups improves alcohol outcomes. The decline in mutual-help group participation therefore likely means poorer treatment outcomes for many patients.

The study can’t show why attendance differs between racial and ethnic groups. Earlier research suggests some people avoid groups they see as designed for White men, or where they have faced mistrust or discrimination. But studies finding higher participation among Black clients in some urban areas suggest the key barrier may be less the groups themselves than being a demographic minority at the meetings available to them.

The researchers caution that simply expanding standard programs that encourage 12-step participation may not close these gaps, because such programs have not proven especially effective for racial and ethnic minority participants. They call for research into the barriers these groups face and for culturally tailored approaches, such as more Spanish-language meetings, referrals to culturally adapted groups and connections to peer support specialists and recovery community centers.

About the data

The data come from the Substance Abuse and Mental Health Services Administration’s Treatment Episode Data Set–Discharges (TEDS-D). They cover publicly funded facilities (about half of all U.S. substance use treatment facilities) and exclude private and most federal programs. The data count treatment episodes, not individuals, but the main findings held when analyses were limited to first-time patients.

Study:

Zemore, SE, Rosen, EM, Mulia, N, Rodriguez Borja, I, Karriker-Jaffe, KJ. Race/ethnicity and Other Sociodemographic Correlates of Weekly Mutual-Help Group Attendance During Alcohol Treatment in 2015–2022 National Data. J Stud Alcohol Drugs. 2026 November. https://doi.org/10.15288/jsad.26-00104.

Funding: This research was supported by the NIH National Institute on Alcohol Abuse and Alcoholism (R01AA029812, R01AA027920).

###

The Alcohol Research Group (ARG), a program of the Public Health Institute, is a nonprofit research group with more than 65 years of experience studying alcohol and other drug use and their consequences for individuals, families and communities. Based in Berkeley, California, ARG's multidisciplinary team of scientists, methodologists and public health researchers works to generate evidence that drives meaningful change in policy, services, and public understanding. For more information, visit www.arg.org.

The Public Health Institute is a national independent nonprofit that advances the health and wellbeing of all people and communities, particularly those facing the greatest barriers. Rooted in 60 years of public health experience, PHI moves ideas into action through practice, policy, research and community engagement, while serving as the operational backbone for more than 40 programs delivering evidence-based solutions to complex health challenges. For more information, visit www.phi.org.

RTI International is an independent scientific research institute dedicated to improving the human condition. Our vision is to address the world's most critical problems with technical and science-based solutions in pursuit of a better future. Clients rely on us to answer questions that demand an objective and multidisciplinary approach—one that integrates expertise across social, statistical, data, and laboratory sciences, engineering, and other technical disciplines to solve the world’s most challenging problems. For more information, visit www.rti.org.

 

New way for hackers to target bank accounts discovered by UMass Amherst researchers



During a bank-to-bank transfer, money can be deposited into one bank account without ever being withdrawn from the other account




University of Massachusetts Amherst

Taqi Raza

image: 

Taqi Raza, Assistant Professor, Electrical and Computer Engineering at UMass Amherst

view more 

Credit: UMass Amherst






AMHERST, Mass. — Researchers at the University of Massachusetts Amherst have discovered how, during a bank-to-bank transfer, money can be deposited into one bank account without the funds ever leaving the second bank. This new class of cyberthreat is called Credit-Without-Debit (CWD) attacks and the researchers have demonstrated three ways hackers can use it to steal money.

 

“Our goal is to strengthen financial security,” says Taqi Raza, assistant professor in the Riccio College of Engineering at UMass Amherst. “The way we strengthen it is we find the loopholes and tell the banks, ‘okay, fix this problem.’”

 

Unlike previous research by this group, which focused on credit card and digital wallet vulnerabilities, this new study examines debit transactions, or ACH transactions. This system permanently moves money between banks—as opposed to creating a debt—and enables any bank-to-bank transaction, including direct deposit of salaries, bill payments (including credit card bills), mortgage services, and withdrawals from digital wallets and various payment apps.

 

The way an ACH transaction should work is that transfers are processed through multi-stage interbank clearing and settlement. Consider the situation where a person has bank accounts at two different banks and wants to move money from Bank A into Bank B. The person initiates the transfer from Bank B, which pulls the funds from the account at Bank A.

 

The transaction is authorized when Bank B recognizes that Bank A has a valid routing and account number and then authorizes the transaction. There is a two-day payment processing window for Bank A to transfer the money, so to expedite access to funds, some banks provisionally post funds in their customer’s account (in this example, Bank B posts these funds). Bank A checks that its customer’s account has sufficient funds, that there is valid authorization, and that there are no account holds, and then sends the money to Bank B. The interbank settlement is complete.

 

However, in a CWD attack, the attacker finds a victim with a frozen account, meaning the account can receive money but not send it. They link the victim’s account to their own and then pull the victim’s funds into their own account. The attacker’s bank posts the provisional funds. However, because the victim’s bank is frozen, the settlement payment is never made. In the two days it takes for this payment to be processed and rejected, the attacker has withdrawn the provisional funds, resulting in a direct financial loss to the bank. The researchers have created an interactive graphic to demonstrate how this attack works.

 

“The key root cause of all these attacks is that banks have treated distinct security checkpoints as interchangeable,” explains Raza.

 

Once they discovered the vulnerabilities, the researchers coordinated with affected banks to resolve the issues, which have not been reproducible since. The team presented their paper on the work today at the ACM Conference on Computer and Communications Security (CCS), 2026.

 

The researchers found two additional instances where this credit-without-debit logic resulted in successful attacks. One attack used microdeposits from third-party payment apps used to verify bank account ownership; the other paid off a credit card bill using a frozen account.

 

“We focused on what the user can observe and what they can influence from their banking apps,” says Ali Malik, Ph.D. candidate in electrical and computer science at UMass Amherst. “They can initiate payments, they can cancel those at random moments, they can observe the progress of that transaction as it moves from processing, to settle, to cancel, to rejected. It’s simple stuff. We defined our threat model around that.”

 

Even with these vulnerabilities discovered, Raza recommends that consumers always follow security best practices, such as enabling two-factor authentication, monitoring bank statements for fraudulent activity and turning on text, email, or banking app notifications whenever a debit or credit happens on an account.

 

Thousands of deceptive ads evade Google's safeguards, study finds



AI tool uncovers deceptive Google ads displayed more than 100 million times, exposing gaps in detection and enforcement




NYU Tandon School of Engineering






Thousands of deceptive advertisements have slipped through Google's advertising safeguards, exposing internet users to fake virus warnings, misleading software promises and potentially dangerous downloads, according to new research from NYU Tandon and Radboud University in the Netherlands.

Using an AI tool they developed called AdLens, the researchers examined nearly 189,000 versions of software-related advertisements from Google's publicly accessible Ads Transparency Center and identified thousands of deceptive ads that collectively generated more than 100 million impressions across European markets.  

Some of the Google ads it identified remained active for years, while others continued running even after being reported to the company.

In one case, Google removed an advertisement linked to a known malware operation but left 41 others directing users to the same website. The findings raise questions about how effectively Google detects harmful advertising and whether its enforcement systems address entire campaigns rather than individual ads.

“Google has systems in place to screen advertisements, but we found thousands of deceptive ads that still reached users, sometimes for years,” said Ritik Roongta, NYU Tandon Computer Science and Engineering (CSE) PhD student and lead author of the study. “Even when we reported ads linked to known malware campaigns, removing one didn't necessarily stop dozens of others promoting the same threat. We shouldn't have to rely entirely on advertising platforms to tell us whether their safeguards are working.”

The researcher team’s study, AdLens: Efficient Detection of Deceptive Software Ads, will be presented on November 18 at the 2026 ACM Conference on Computer and Communications Security (CCS) in The Hague, Netherlands. The research was overseen by Rachel Greenstadt, CSE professor and faculty member of the NYU Center for Cybersecurity and Gunes Acar, associate professor at Radboud University.

AdLens identified three major categories of deceptive advertising:

  • Scareware Ads: Ads with fake warnings include those claiming that users' devices were infected, hacked or compromised generated more than 16 million impressions. The researchers identified 238 such ads within the 189k ads examined.
  • Misleading software promises: The researchers identified 3,346 ads making false or unrealistic claims about what software could do, including recovering photos deleted years earlier. These ads generated nearly 90 million impressions.
  • Deceptive designs: The researchers identified 258 ads that concealed the advertiser's identity or what was being promoted, sometimes displaying little more than a button labeled "Continue" or "Access Now." One such ad generated 2.1 million impressions.

One advertisement promising to reveal anyone's GPS location using only a phone number generated 14.5 million impressions and remained active for more than two years.

“These ads are only the tip of the iceberg, as the ad transparency platform hosts more than two billion ads of which only a small portion was analysed,” said Roongta. “The high incident of deception within this small sample itself highlights the magnitude of these scams being propagated to the users.”

The researchers also examined apps promoted through deceptive advertisements. Of 147 identified apps, 34 were no longer available in app stores even though their advertisements remained active.

AdLens uses publicly available AI models (small language models or SLMs) to identify suspicious advertising efficiently. It first searches for language resembling known deceptive tactics, then examines the text and images of likely offenders. By concentrating intensive analysis on a smaller group of ads, the system reduces the computing resources needed for large-scale investigations.

The study builds on earlier research by Roongta and Greenstadt, which found that users of Adblock Plus's "Acceptable Ads" feature encountered 13.6% more problematic advertisements than users browsing without ad blockers. Together, the studies reveal shortcomings in systems intended to improve the online advertising experience.

The researchers have made AdLens publicly available so regulators, consumer protection organizations and independent researchers can investigate advertising practices without relying exclusively on platforms' own assessments. Although this study focused on Google, the technology could be adapted to examine other advertising platforms, including Meta and TikTok.

The researchers caution that ad impressions measure exposure, not confirmed financial losses or malware infections. Nevertheless, their findings demonstrate how deceptive advertising can persist despite existing safeguards and underscore the need for stronger enforcement and independent oversight.

In addition to Roongta and Greenstadt, the authors of the AdLens paper are Marwan Adnan Darwish, Masoud Poorghaffar Aghdam, and Gunes Acar, all of Radboud University. The research was supported by the National Science Foundation and the Netherlands Organisation for Scientific Research.