A small fix could make a big difference in Tijuana River pollution: When will it happen?

Stannard’s story is all too common.

Over the last year, I’ve interviewed more than 20 patients, healthcare providers, experts and advocates about the privacy forms they must sign to get care at their providers’ offices.

Time and again I was told the same thing: Across the country, from large hospital systems to small, private clinics, patients are being asked to sign waivers blindly without knowing exactly what they’re signing.

When patients ask to see more, staff usually don’t have an easy way to show them. When patients do get the forms, it tells them all the ways their medical data will be shared and reused, and some of the ways patients can refuse. But electronic systems make it impossible to opt out on the spot, requiring follow up emails.

Records sharing between unaffiliated providers through these networks can benefit patients by making their scattered records more visible to the provider who is treating them. 

But it can also harm patients.

To experience more of what patients have to deal with and test whether they’re able to successfully get the information they need, say no, or opt-out of having their data shared, I checked out over a dozen health care systems myself by registering and going to appointments in Iowa, New Jersey, New York, Ohio, Oregon, South Carolina and Virginia.

One telehealth appointment with a provider showed me how easily dark patterns force patients to share their data with big healthcare networks, even when the privacy form they’re signing explicitly says they can opt-out. 

In October 2025, I booked a telehealth appointment with a women’s health clinic in Virginia, after a source was frustrated with the clinic’s check-in process. During registration, I was asked to sign their notice of privacy practices. It’s the same type of form that Stannard never got, but was asked to say she did.

The notice told me that I was giving them permission to let my physician share my health data with a health information exchange, a network that allows providers to search my medical records, like lab results or medical history, from other health organizations when they treat me. These networks can be regional, state-wide or national in reach. The privacy notice says that by signing the form, “you agree to have your medical information shared.” 

It also says I have two other choices:

• Say no by following instructions on the opt-out form, but there’s no link to the form. 
• Say yes now and kick off the opt-out process later by sending an email. An email address is provided.

But when I got to the end of the privacy notice, I wasn’t allowed to say no. I had only one choice: “I accept.” After that, there’s a spot to type my name “to accept the policy,” check a box that I understand that I’m electronically signing, and a big button to “Continue.”

I ignored the accept button and tried clicking “Continue.” An error message told me I couldn’t move forward unless I hit “I accept.”

I was at a crossroads. The privacy notice literally describes “Say No Thanks” as a choice, but doesn’t let me pick it.

At this point, most of the patients I’ve interviewed would probably click “I accept” and move on, even if they wanted to keep their information private. But I was researching what patients have to do for healthcare systems to honor their wishes around consent and privacy, so I stopped filling out the form.

Instead, I emailed the address on the privacy notice. I was surprised that an employee got back to me that day, shared the opt-out request form, and confirmed that “registration is required to opt-in.” She also told me her company, which manages this consent process for the information exchange, will process my opt-out after I sign it and they’re able to process it. The risk is that they might not do it before my appointment. I emailed her back and asked what we should do about this, since the original privacy notice says, “Please note, your opt-out does not affect health information that was disclosed through HIE [health information exchanges] prior to the time that you opted out.” How could we make sure none of my information is shared? 

The next day, she replied that her company would proactively opt me out of the information exchange, that I should still complete the opt-out form she sent me, and that “You should now be able to complete your check-in, and the setting will remain unchanged.”

When I went back to check in for my appointment, I clicked “I accept,” because the health services company assured me nothing will change. Just to be safe, I wrote “I opt out of HIE” and my initials, “AR” into the box where I’m supposed to write my name.

When I wrote to a manager of the women’s clinic about this, they stood by Privia’s process and said that Privia makes themselves available for patients who want to opt-out.

“This is a dark pattern,” said Lior Strahilevitz, a legal scholar at the University of Chicago who has published papers on privacy and dark patterns and teaches health law. In fact, Strahilevitz sees multiple dark patterns in the patient registration process I went through.

One is called an “obstruction dark pattern,” which means the design makes it harder for patients to make any choice except the one healthcare providers want. 

Another dark pattern was “visual interference” where the interface makes it hard on the patient. “The patient’s going to have to face inordinate burdens in order to make an autonomous choice,” he said, because they will need to go “outside the user interface, outside the screens, in order to exercise your opt-out rights.”

Lucia Savage, former chief privacy officer at the federal health IT office, called the Office of the National Coordinator for Health IT, said that problems like this can happen when people carelessly put physical forms online. “This isn’t really a design at all,” she said. “This is just a bunch of paper pasted onto a web page. Could you even really call it design?”

So, is all of this legal?

Legal experts point out that only one element of the check-in process violates the spirit of health privacy law, and it’s not the part I expected.

In Virginia, where I had my appointment, it’s legal for providers to opt patients in at registration and give them a way to opt-out later.

Some states, like Florida and New York, require providers to get a patient’s explicit consent before they can share or access a patient’s data from information exchanges. Other states, like Arizona and Maryland have laws that allow data-sharing through health information exchanges by default, as long as providers tell patients and give them a way to opt-out. Some states have not passed any additional regulations, which means they follow the federal baseline. Federally, under the Health Insurance Portability and Accountability Act (HIPAA), sharing a patients’ data in a health exchange is legal.

According to Sarah Jaromin, a health policy specialist at the National Conference of State Legislatures, in Virginia, there is no current state policy with explicit opt-in or opt-out requirements.

Craig Konnoth, a law professor at the University of Virginia who specializes in health and civil rights looked at the privacy notice I was asked to accept. “You have the choice as to whether your data is going to be used. In this particular situation, ‘we are going to use your data until you file in the opt-out paperwork’ — then that’s actually kosher,” he said. 

What experts say violates the spirit of the law, however, is requiring that patients sign the privacy notice itself. 

When I was checking in, the privacy notice forced me to add my signature and click “I accept” before I could click “Continue.”

“What becomes problematic for me is that you can’t actually proceed. The design forces you to do something that the HIPAA privacy rule does not require you to do,” said Stacey Tovino, a professor who teaches HIPAA privacy law at the University of Oklahoma College of Law. (Full disclosure: As a part of my role as Director of Sociotechnical Research at The Markup and CalMatters, I am combining a broader journalistic investigation with a small ethnographic research studying on digital patient intake procedures, The Markup paid Tovino to consult on the HIPAA implications of my findings, but she did not participate in data-collection or editorial decision-making.)

“Nothing in HIPAA requires them to make you sign the notice,” said Tovino. “If they don’t obtain the signature they simply have to document why they didn’t get it.”

There’s an important nuance here. At a doctor’s office, patients usually have to sign and give consent to treatment and financial responsibility policies before they can actually get medical care. But when it comes to privacy notices, HIPAA only requires healthcare providers to ask that patients acknowledge receiving it. Patients should be able to ignore it. 

Many of the privacy-focused patients I interviewed, including those who also work as doctors and nurses, deliberately decline to sign a notice of privacy practices if it contains terms they disagree with. But when modern check-in technology refuses to let a patient move forward without agreeing to the notice of privacy practices, is that legal?

Emily Hilliard, press secretary at the U.S. Department of Health and Human Services (HHS), confirmed that the HIPAA privacy rule does not require providers to get a patient’s consent to their privacy notice, but it also does not “prohibit covered entities from requiring individuals to acknowledge, or agree to the terms of, an NPP.”

In other words, requiring patients to agree to a privacy notice before getting treatment is legal.

“Likely because HHS never envisioned this happening, HIPAA does not explicitly prohibit a covered entity from requiring an acknowledgement of receipt of the notice of privacy practices as a condition of treatment,” said Adam Greene, a partner at the law firm Davis Wright Tremaine who focuses on health information, privacy and security.

“HHS has heard about widespread problems with the acknowledgment of receipt of the notice of privacy practices becoming an obstacle to patient care and a cause of confusion,” he said. “In 2021, they issued a proposed rule that, amongst other things, proposed deleting the requirement for an acknowledgment of receipt of the notice of privacy practices.” The rule was never finalized, but it is back on the agenda this year. 

Stannard confirmed that at HHS, “we are in the process of finalizing the rule which includes some additional requirements for the notice of privacy practices.”

The current proposed rule includes, “Eliminating the requirement to obtain an individual’s written acknowledgment of receipt of a direct treatment provider’s Notice of Privacy Practices.”

Experts say patients should be able to opt out immediately — not eventually

Legal experts say that regulators can fix this problem with one fell swoop: make it a rule that companies must let patients opt-out right away, at the same moment they’re notified that they can.

“Amend these [federal] regulations to say covered entities shall not impose an undue burden on people trying to opt out. Covered entities shall not make it functionally problematic. Covered entities shall not, in registration documents, force people to proceed, thus waiving their right to opt out at the earliest possible time,” Tovino said.

She suggested that when a company notifies someone of their right to opt out, the next sentence should include a link to do so.

Savage agreed that this change would “absolutely” be a substantial intervention. “I believe that’s something OCR [Office of Civil Rights at HHS] could do in regulations.”

At the same event where Stannard shared that her eye doctor asked her to acknowledge a privacy notice she never got, I asked her, “Would updating the privacy rule to require a live link when patients make a choice to opt out or into sharing their information be empowering to Americans as individual patients?” She’d just spoken about U.S. Health Secretary Robert F. Kennedy Jr.’ s agenda “to empower individuals with their own health information.”

“That’s an interesting idea,” Stannard responded. “I don’t remember if we’ve considered it before. It’s certainly something that we could consider going forward.” 

One registration form, but a cocktail of technology companies

Navigating the dark patterns in the check-in process was difficult. What I’ve learned however, is that it’s hard to know who picked that interface to use with patients. Did it come from the clinic or the sprawl of vendors that health facilities have come to rely on? 

Private clinics often partner with multiple outside companies (vendors covered by HIPAA) to get technology and administrative support. My appointment involved three different companies: 

• The mobile link I received to check-in for my appointment comes from a company named Phreesia, which handles patient-facing software, like consents, medical screening surveys and payment. When a patient clicks through those consent forms in the U.S., it goes through Phreesia every 1 in 6 patient visits.
  
• The clinic had joined Privia Health, which handles management services for nearly 5,000 providers across 15 states, which affect 5.2 millions patients, according to a 2025 press release. The privacy notice I struggled with sent me to Privia’s medical records office to opt out. Phreesia’s logo was also on the copy of my forms that the clinic emailed me. 
  
• Finally, for my second telehealth appointment six months later, the clinic sent me a link with the name of another vendor, “athenahealth,” in it. The clinic had replaced Phreesia with athenahealth entirely.

“Unless you’re a really giant system,” said Savage, “you don’t have internal expertise on how to do this. So you buy it. You buy what’s plug-and-play and what’s affordable.” 

The Markup and CalMatters asked all three companies who was responsible for the design of the patient registration interface, and no company gave us a clear answer.

Privia: “Privia is committed to the privacy and security rights of our patients’ information and to ensuring we comply with all regulatory requirements regarding our use of that information,” said Robert Borchert, senior vice president of investor and corporate communications at Privia Health.

athenahealth: “athenahealth provides technology that healthcare providers use to manage patient registration and clinical workflows … configured according to each provider’s requirements and applicable law,” read a statement from athenahealth, provided by Nikki D’Addario, senior public relations manager.

Phreesia: “It is the provider’s form and they determine the content and interface options,” said Dori Zweig Young, Phreesia spokesperson.

None of the companies responded to detailed written questions about how much control clinics have over the interface.

A blindspot for regulators and how it can be fixed

Outside of healthcare, regulators, like the Federal Trade Commission (FTC), the Consumer Financial Protection Bureau (CFPB) and multiple state attorneys general and agencies, have called dark patterns manipulative or abusive tactics that confuse consumers about their privacy choices or lock consumers into paying for services (like the famous Amazon Prime case). Researchers consistently find that people want more control over the context of how their data is shared, and that patients are least comfortable handing over blanket access with broad, open consents, even if they are largely willing to share it for specific uses.

Strahilevitz explained, however, that agencies like the FTC and CFPB, which have been the most active on regulating dark patterns, regulate privacy within their zones, and only occasionally take on boundary cases.

“Health privacy, for the most part, is going to be primarily addressed by HIPAA and Health and Human Services rather than the FTC Act and the Federal Trade Commission,” she said.
“There are limits on [the commissions] ability to protect patient privacy because that’s basically another entity’s job.”

Green and Savage both agreed that the Federal Trade Commission has jurisdiction to enforce against dark patterns as unfair or deceptive practices in for-profit healthcare entities. The clinic I went to, like hundreds of thousands across the country, is for-profit. 

But HHS has a broader mandate to regulate healthcare organizations, including non-profit hospitals. 

For example, Strahilevitz said, in consumer finance, regulators at the Consumer Financial Protection Bureau treat a practice as unfair or deceptive when a consumer cannot reasonably avoid the resulting injury. Just as hard-to-cancel online subscriptions force people to pay more, maze-like opt-out structures force patients to pay with their data by default.

Strahilevitz said this provides a framework for thinking about privacy injuries in healthcare. An information exchange could serve as a clearing house for information about a patient’s abortion, which has a clear potential for injury if that information becomes known in a state where abortion treatments are criminalized.

“In other privacy contexts, the courts have said where it’s literally possible to opt out of something but, practically quite difficult, unduly onerous, then we’re not going to treat that as creating an opt-out right,” he said. 

Savage sees more opportunities in carrots than sticks to get to best practices. She argued that the government could invest in good interface design that’s open source and available for anyone to use, and the federal health IT office, where she used to work, could create competitions focused on improving the technical tools that providers buy and use.

If the big technology vendors that independent clinics are already using make these changes, it could affect millions of patients.

State regulation is another possible solution. Strahilevitz said that scrutiny of dark patterns is spreading as states, like California, and regulatory agencies, like the FTC, seek to reign in unfair or deceptive practices through the simple intervention that it should be as easy to cancel as it is to subscribe, with one click.

“I hope that at some point, we’ll get to a point where symmetry of choice is the law of the land, not only with respect to consumer privacy in some states, but to these kinds of medical privacy or financial privacy or other contexts,” he said.

California’s community colleges represent the largest higher education system in the country — more than 2 million students, or 60 times the undergraduate population of UC Berkeley. But walking around a community college campus, it’s often hard to tell. 

Since the COVID-19 pandemic, cafeterias and local coffee shops are quieter, fewer students are sitting on the quad and, with less foot traffic, the grass is lush. Even after campuses returned to in-person classes, many students are still working from their dining room table: About 40% of all community college classes are online, according to Melissa Villarin, a spokesperson for the California Community Colleges Chancellor’s Office. 

The state’s community colleges are funded based largely on the number of students they enroll, and since students prefer online courses, there’s an incentive for schools to expand them.

Ask students or professors about the merits of online education, and they’ll often say it’s more accessible, especially for students who have kids or are working a full-time job. The same argument is often true at the University of California and California State University campuses, which offer considerably more online courses than before the pandemic, though far fewer than the community colleges. 

Ask students or professors about the problems of online education, and they’ll point to any number of familiar complaints: a lack of engagement, a sense of loneliness, impersonal lectures, and the temptation to move the Zoom window aside and click on something else. In online classrooms where the majority of students keep their cameras off, bots and scammers have become a systemwide problem: they use AI and other algorithms to mimic real students, submit assignments and steal financial aid. Even real students are using AI to submit online assignments, while teachers are using it to grade.

Researchers say it’s hard to know how the quality of online education compares to in-person courses because it’s subjective and because of the wide diversity of courses and  teaching methods. 

In Lupe Archundia’s microeconomics class at San Joaquin Delta College in Stockton, all the lectures were pre-recorded, in some cases more than a decade ago. The professor gives students the answers to the quizzes — before they take the test — and all the quizzes are in a multiple-choice format that a computer grades. 

“I am a 39-year old woman,” Archundia said. “It’s not like I just finished high school and I want easy test answers.”

Archundia has two kids and a full-time job as a secretary, so she studies in the evenings, turning her dining room table into a standing desk with the help of a few cardboard boxes. She wants a bachelor’s degree to help her move up in her career.

In the beginning of the course, she said she would study for three hours before completing each quiz, but once she discovered the professor had made the answers available, she started cutting corners. She said there are still certain concepts, such as elasticity, that she doesn’t fully understand, even though she aced the online exam. 

She feels conflicted about it. “I’m responsible, too,” she said. 

What the research does — or doesn’t — say

The research into online education is generally inconclusive. One 2025 study found that students consistently perform worse in online classes than in-person ones, though the gap is decreasing. Online courses also make it easier for students to hold a job while in school and complete their degree in the long term, said Di Xu, a professor at UC Irvine’s School of Education. 

When asked about students’ concerns with online education, Alex Breitler, a spokesperson for Delta College, said these classes expand “access to higher education for working adults, parents, caregivers, and other students balancing significant responsibilities,” including many students who “simply would not be able to pursue college without online options.”

Delta is not alone — the idea that online courses increase access is a common refrain among college officials. Xu pointed to one empirical study of an online master’s program at Georgia Tech that proved this point, though the  students are very different from those at California’s community colleges, where many are seeking short-term career training or an associate degree.

What researchers do know is that online education has inherent challenges. It requires “self-directed learning skills,” including a “very high level of self-time management,” said Xu. “In an in-person environment interaction happens naturally,” she said. “But in an online environment, especially asynchronous, that opportunity needs to be embedded. Otherwise, the student will feel very lonely.” 

The majority of online classes at California’s community colleges are asynchronous, meaning that the content is all pre-recorded and students can study at their own convenience. Students prefer asynchronous classes too, even compared to online courses where the instructor is live,  according to a survey by the RP Group, an education research nonprofit.

Archundia said she always opts for in-person classes but there are few available, especially for the English classes she wants to take and during the evening hours that she’s available. Her dream is to become a writer, and she wants to switch her major to English, instead of her current major, business administration, though she isn’t sure what classes are necessary to make that happen. 

In April, when she reached out to a college counselor for help selecting classes, the next available appointment was about three weeks later. Archundia still hasn’t been able to find an appointment that works with her work schedule.

One-on-one advising and support structures, such as guidance counselors, are essential for online students, said Rebecca Ruan-O’Shaughnessy, the director of program and strategy at College Futures Foundation and a former executive at the California Community Colleges Chancellor’s Office — but schools also need to adapt. 

Online courses are fundamentally different, and schools need to redesign their courses, not just retrofit them, she said. She pointed to some programs that have new and promising approaches to online education, such as shortening the length of the class or trying to integrate adults’ work experience given so many online students have a full-time job. 

“That is the difficult part for community colleges and other institutions,” Ruan-O’Shaughnessy said. “Frankly, they don’t have the incentive to do that level of work, because that’s a lot of work.”

Breitler, with Delta College, acknowledged that counseling appointments are often booked “weeks in advance” because of high demand. He said the college is trying new solutions, such as letting students submit questions to counselors online and creating drop-in hours where an appointment isn’t needed. 

Remedial education in foreign languages

Cyndi Cunningham enrolled at Palomar College in San Marcos, on the northern edge of San Diego County, in 2022, after the pandemic forced her local shopping mall to close temporarily, making her longtime retail job suddenly seem precarious. Starting college for the first time, she was taking general education and introductory courses, mostly online, and struggled to pay attention and manage her time. “I only ended up taking one class in person per semester — not because I didn’t want to take in-person classes — but because I couldn’t find them,” she said. “I felt like I wasn’t learning; I was just kind of doing tasks.”

She saw professors cutting corners too: Two of her classes in Chicano Studies were taught by the same professor and she once noticed he was using the exact same lecture in both classes. 

Cunningham has since transferred from community college to Cal State San Marcos, where she’s majoring in ethnic studies and plans to become a high school teacher. “Even engaging with other students is so much different in person than on a discussion board,” she said. “I realized more how much of a disservice the online classes did.”

To an extent, online classes can save costs for colleges because they don’t require a physical space and they can enroll many more students, said Xu. But she said adding support systems — such as specialized counseling for students or professional development for faculty — can create additional expenses. Online education “has the potential to save a lot of cost,” she said, but only if colleges are “willing to sacrifice a lot of the quality elements that are important for students.” 

Foreign language courses are particularly costly for universities, said Julia Simon, a professor of French at UC Davis and the chair of a task force on languages for the university. Language courses are typically small, meet regularly, and many less popular languages enroll only a handful of students. Facing a structural budget deficit, the university recently asked her task force to develop a plan for slashing courses in the event of cuts. 

Meanwhile, she said both the nearby community colleges and the UC system are expanding online foreign language classes, which can operate at a larger scale. Sacramento City College, for instance, is offering four French classes in fall 2026 — all of them are online and fully asynchronous. 

“It’s an enormous problem,” she said. In her view, the students who take online courses lack the same opportunities to practice their speaking and miss out on vital cultural lessons that don’t fit in a strict language-learning curriculum. Once they enter UC Davis, they’re unprepared, she said. “We can’t make them repeat courses they’ve already had.” 

She said she’s considering creating a set of conversation classes that would amount to remedial education. 

‘It all depends on the professor’

California legislators and education officials have poured millions into improving online education since the pandemic and have introduced new rules meant to encourage more interaction between faculty and students. All across the state, faculty routinely train on ways to improve their online instruction, and colleges have hired staff members to help with online course design and scheduling.

But the 2024 survey by the RP Group found that among faculty who had taught at least one online course, the majority still preferred in-person instruction. 

Tina Rocha’s creative writing professor at San Joaquin Delta College recently took a sabbatical, learning how to improve teaching for people with learning disabilities. It paid off, said Rocha, who is 55 and started college in 2024 after recovering from three back-to-back strokes in 2020. Because of her disability, she occasionally needs reminders from the instructor to submit assignments. Sometimes she asks for accommodations to avoid certain noises or lights that distort her vision and make her twitch, she said, but her professor is understanding and accommodating. Online education can be a “wonderful alternative,” she said.

Rocha studies every night at her dining room table, which is often scattered with her notebooks.  A calendar hangs from her wall, with notes covering every corner of white space, and a white board sits at the entrance to her home, listing out in color-coded lines each of the week’s responsibilities. 

“It all depends on the professor,” she said. Her online film class this semester has been much worse than her creative writing course, she said. The film professor has a lava lamp in the background that reflects psychedelic patterns on the ceiling. When Rocha asked him to turn it off, he said he tried but was unable to, without offering an explanation. Now, to prevent symptoms, she places a sticky note on the screen whenever the professor starts talking. 

Rocha said she tried to switch to an in-person film class but was too late. Only online classes were available.

---

We’re bringing our voter guide to life through VotingMatters events across California this month, in collaboration with on-the-ground partners: Local news organizations, colleges and nonprofits. Our next event is this evening in Merced and Fresno and Wednesday in Modesto. Plus, we have a DIY kit to host your own event.

---

Other Stories You Should Know

---

Worst threat at Garden Grove chemical tank is over, officials say

CA courts are putting AI to work. Any objection?

---

---

Other things worth your time:

Some stories may require a subscription to read.

---

Two of California’s largest courts are testing an AI tool that can draft orders and produce research memos. 

Judges so far are using it primarily for civil cases, but documents obtained by CalMatters indicate the possibility of expanded applications in criminal cases, where people’s freedom and access to justice are on the line. 

The Los Angeles County Superior Court began a pilot program in February to test a tool created by the company Learned Hand. Other courts may follow, according to Learned Hand founder and chief executive officer Shlomo Klapper.

Learned Hand uses a combination of language models from Anthropic, OpenAI and Google to act as an AI clerk for judges. The company says it tests for bias and accuracy, but it has not yet published results. 

In Riverside County, which has a $10,000 agreement with the company to test the program, civil and probate attorneys are primarily using the tool to draft research memos that help judges reach their decisions. It’s typical for research attorneys to assist judges as they review cases.  

Los Angeles County Superior Court has a roughly $314,000 contract that includes a roadmap to test the tool’s use in criminal, family and probate divisions. Officials would not describe in detail to CalMatters the criteria they’re using to evaluate whether use of the tool can safely expand to criminal and family courts, where the stakes are often much higher than in civil cases. 

One judge who spoke to CalMatters on condition of anonymity due to judicial rules of conduct was alarmed when their colleagues at a recent luncheon said the technology could be used one day to evaluate appeals from people who believe their conviction or sentence was tainted by racial bias. California courts are handling a wave of those claims after lawmakers passed the Racial Justice Act in 2020. 

“I think it is outrageous,” said the Los Angeles County Superior Court judge. “AI cannot and never will be able to replace human judgment in evaluating complex social dynamics. Ultimately, that will erode the public’s confidence in the competence and fairness of the judiciary.”

A majority of California’s superior courts now have generative AI use policies, according to documents obtained by CalMatters via public records requests, which they were required to create by the state Judicial Council before using the technology. Roughly a dozen of the 51 courts that have responded to CalMatters’ requests said they are using AI-powered tools from LexisNexis, Thomson Reuters, and Microsoft’s Copilot.

Use of AI in courts has been controversial because of the propensity of AI models to cite falsehoods and to produce sycophantic text. Models from major companies like Google and Anthropic can reduce critical thinking and brain activity, according to a 2025 MIT study.

Language model hallucinations have already made it into the judicial system. Researcher Damien Charlotin has documented hundreds of instances of litigants, lawyers, and judges making mistakes when using AI to do their jobs including nearly 90 cases in state or federal courts based in California since August 2024. 

Last fall, a Los Angeles-based lawyer received a historic $10,000 fine for citing cases that don’t exist, and earlier this month the Sacramento Bee reported that use of AI led to errors in four cases handled by prosecutors in Nevada County. Most of these cases involve lawyers or people who are representing themselves in court, but UCLA Law School professors predict that more judges will make AI-fueled mistakes in the future. In recent months, the U.S. Senate investigated federal judges in Mississippi and New Jersey for drafting decisions with generative AI that had serious factual errors. 

Klapper, who previously worked as a clerk for a federal appeals court and for surveillance technology company Palantir, said the judiciary needs AI in order to reduce backlogs and increase efficiency.

“Could we hire more people?” he told CalMatters. “Maybe, but it’s not going to keep pace with the exponential increase that’s coming, nor is it going to be able to adequately solve the crisis of today. I think the only solution is to give every single judge and staff attorney their own AI clerk.” 

Klapper said he’s aiming to combine the best parts of what human judges can do with the best parts of what machines bring to bear. 

“I’m not saying all machines aren’t biased,” he said. “I’m not saying my machine isn’t even biased. I’m saying we can test it and people have tested it. And that is the benefit over humans.” 

Generative AI use policies for the Los Angeles and Riverside County superior courts only require disclosure if a motion, decision, or other document is written entirely with generative AI. 

Both courts refused to say whether plaintiffs are aware that the tool is being tested on their cases. In a statement to CalMatters, a spokesperson for the Los Angeles County Superior Court said testing is done on motions that have already been decided, separate from live case environments. However, the contract allows for testing on live cases.

“It is important to note that even with successful evaluation and thorough testing, the Court remains several months, if not years, away from implementing this type of tool,” said the spokesperson. 

The contract allows the tool to be used for two critical motions in the criminal division: A motion to suppress, which is designed to determine what type of evidence the prosecution is allowed to present at trial, and motions for post conviction relief, which are filed by people who have already been convicted and want another shot at freedom. 

That’s the “greatest concern” for Los Angeles County District Attorney Nathan Hochman. When he reviewed the contract, he referred to the motions as “two incredibly important motions in the criminal justice system.”

“When you’re dealing with someone’s liberty — as opposed to in the civil setting, which is everything other than liberty — the stakes couldn’t be higher,” said Hochman. “I don’t want to take the chance, particularly in a criminal case, that AI happens to get it wrong. And now someone’s constitutional rights have been infringed. Someone has gone to prison who shouldn’t have, or on the flip side, that somehow someone gets off.”

‘An extremely perilous road’

In Los Angeles, some judges first heard about the new Learned Hand contract during a March presentation by Superior Court Judges Yvette Verastegui and Olivia Rosales. They lead the criminal branch and visit courthouses throughout the county as part of an annual roadshow, where they update judges on court operations, discuss workload and field questions. During a luncheon, Verastegui and Rosales said the tool could be used to assist with Racial Justice Act petitions in the future. 

California’s Racial Justice Act allows people to challenge a criminal conviction or sentence that they believe was based upon racial bias. Petitions are filed directly to the court from people in state prison. If a case is found to have merit, the process includes appointing legal counsel, filing briefs and setting evidentiary hearings before a judge would decide whether to grant the petition. 

That process could look different with a tool like Learned Hand. Verastegui and Rosales explained that, following an incarcerated person’s petition, the tool could generate tentative decisions for judges to consider in denying or advancing cases to the next stages, according to one judge who attended the luncheon. 

“The concern, of course, that I have is that the courts will utilize that as a reference point and then get stuck to that initial analysis,” said the judge. “It’s an extremely perilous road to go down. Putting aside the inaccuracy, which will be a significant concern, it dehumanizes the whole process. It does not treat people as individuals with lived experiences. It essentially reimposes a one-size-fits-all style of justice.”

A second Los Angeles Superior Court judge who spoke with CalMatters on the condition of anonymity remembered the presentation and said they would not trust nor use the tool to summarize a Racial Justice Act petition.

AI can replicate or intensify patterns contained in the data used to make a model, including human biases. Large language models have a history of demonstrating race and gender bias, an analysis of predictive policing tech used by LAPD found racial bias, and an analysis of the risk assessment algorithm COMPAS found that it is more likely to label Black people as at risk of committing crimes after incarceration than white people with a similar record. 

Public defenders who spoke with CalMatters echoed those concerns. 

Elizabeth Lashley-Haynes, a deputy public defender at the Los Angeles County Public Defender’s Office, said it would be “highly problematic and bordering on unethical” for a judge to use the tool to review Racial Justice Act petitions, which she described as “incredibly nuanced.”

“They’re like nothing else in the legal system that has ever really been done,” said Lashley-Haynes, who specializes in Racial Justice Act cases. “Words that are used in these cases that have racial undertones or racial meanings are way beyond the realm of anything that artificial intelligence could do.”

In interviews with CalMatters, Klapper and Los Angeles County Superior Court Executive Officer, David Slayton, denied that the court has any plans to use the tool for Racial Justice Act petitions. A spokesperson for the Los Angeles Superior Court later confirmed in an email to CalMatters that the contract permits the tool to be used in such a way “but that possibility has not commenced in any way.” 

Klapper said if they were to build out a Racial Justice Act module, the tool would need to be evaluated for bias and co-developed with the court. 

“The timing very fortuitous, right?” he said. “It’s a very fraught decision, I’m not going to lie…extremely high stakes — a scenario where I understand people might be very concerned. Especially with criminal, I have even more hesitancy, even more guardrails than normal about, because there are liberty interests at stake.”

Extending beyond civil cases

In Los Angeles, six superior court judges and their research attorneys are primarily using the Learned Hand tool to conduct research, summarize motions and assist in drafting tentative rulings, according to Slayton. He says the tool won’t move beyond the civil division “until the court leadership is comfortable.” 

“The court is being very deliberate and careful about how we use technology like this,” he said. “So until we evaluate it and determine that it is effective in those areas, we will not extend it to other areas.” 

The tool will be evaluated on a quarterly basis to determine its future application, Slayton said, but he did not specify what kind of evaluation that entails. In an email to CalMatters, a spokesperson later said that Learned Hand is evaluated “against the same substantive expectations applied to law clerks and research attorneys: accurate legal research, sound analysis, neutral and judge-ready writing, and reliable work product that supports judicial decision-making.”

Los Angeles County Superior Court Judge Samantha Jessner, who chairs the Judicial Technology Advisory Committee, said she was unaware of the possibility that the tool could eventually be used outside of the civil division until recently. Judges are not privy to contract negotiations due to certain ethical limitations, she said. 

“I think we have a duty and obligation to explore whether or not there is a place for artificial intelligence in what we do as a judicial branch and that’s exactly what this pilot is intended to afford us the opportunity to do,” said Jessner.

Riverside County Superior Court signed an agreement with Learned Hand in February. In emails obtained by CalMatters, Klapper proposed to two Riverside County Superior Court executives, Jason Galkin and Sarah Hodgson, that the court use the tool for a common civil court motion and “then expand quickly once we earn our stripes.” He suggested that Hodgson assemble a list of motions and workflows “that generate the most pain,” citing examples that included the Racial Justice Act. 

Roughly two weeks later, Hodgson described the most laborious motions “that want to drive us into retirement,” including discovery motions and attorney fee motions. For criminal cases, the court suggested that Klapper focus on “things with the largest paper records,” citing death penalty habeas petitions and parole revocation.

Since the pilot started, seven civil and probate attorneys have been granted access to the tool. Galkin, the chief executive officer of the Riverside County Superior Court, said they are “kicking the tires on the product” to see what tasks it can do. The tool is not being used to draft tentative rulings, he said. 

“We don’t even know if expansion is likely so there is no set criteria for what expansion might look like or thresholds for that because right now, the core question is: Does this help staff and does it advance what they’re trying to do in their roles?” said Galkin.

As testing is underway, attorneys like Hochman say that use of AI is inevitable, but would be better suited for low-level, repetitive and routine tasks.

“It’s the analysis of the case itself, coupled with the conclusions that will be reached, that I’m very hesitant to trust AI at this point — in large part, because I don’t know all of the inputs that AI is using to make its decision. The only thing I’m 100% sure of is that AI didn’t go to law school,” said Hochman.

Cayla Mihalovich is a California Local News fellow.