Ask the Experts: AI Surveillance and US Immigration Enforcement

Dia Kayyali is a fellow at Tech Policy Press.

Despite myriad legal challenges, errors, and profound moral costs, the Trump administration continues to cancel student visas with no notice and defend its illegal deportations. US citizens aren’t exempt. The Department of Homeland Security (DHS) has already improperly detained citizens and recently sent a mass email threatening citizens and non-citizens alike with deportation. The administration has pledged to increase social media surveillance and the use of AI to flag individuals it deems a threat. In this context, civil rights groups and journalists are sharing “know your rights” materials and lessons on digital security.

I spoke to six experts to shed some light on what we do and don’t know about how the Trump administration may be using AI for immigration enforcement, and how at-risk travelers, visa holders, and citizens can stay safer, including:

1. Faiza Patel, Senior Director for the Brennan Center's Liberty and National Security Program
2. Petra Molnar, Associate Director of the Refugee Law Lab at York University and a Faculty Associate at the Berkman Klein Center for Internet and Society at Harvard University
3. Dave Maass, Director of Investigations, Electronic Frontier Foundation
4. Nathan Freitas, Founder and Director of the Guardian Project
5. Zahra Billoo, Executive Director of Council on American-Islamic Relations, San Francisco Bay Area
6. Raya Sharbain, security trainer

What’s really new and what we don’t know

The use of artificial intelligence and social media surveillance by US immigration authorities is not new. Presently, the Trump administration is largely using technology and programs that were developed and acquired by previous administrations, even as it invests in new tools US civil society organizations have made public a lot ofinformation about the Department of Homeland Security, from social media surveillance to border enforcement technology and tactics including how AI plays a role. Here’s what experts had to say about what may have changed.

Faiza Patel:

Starting with the Obama administration, the federal government has built an extensive infrastructure for agencies to comb online speech. For example, since May 2019, the State Department has collected social media handles from some 14 million visa applicants annually, which are saved indefinitely in government databases. These give the government an easy way to track what foreigners studying or working in the US are saying and map out their social networks.

I would love to know how they decide that online speech is anti-American, antisemitic or pro-Hamas. As I said in an earlier piece, “These are broad and contested terms. Pro-Palestinian and anti-Israel sentiments have often been conflated with anti-Semitism or pro-terrorism, leaving a broad swath of people vulnerable to being caught in an AI-enabled social media net.”

Petra Molnar:

The current Trump administration has ramped up [existing] efforts, heralding the most recent era of using high-risk technologies at the border at the expense of people’s rights and lives. 

It would be important to know where the border surveillance industry is heading in terms of new technologies and the priorities of the administration. Are they more focused on surveillance or data gathering, or using AI to automate parts of the immigration enforcement machinery? Although, as the last few months have already shown us, it looks like the answer may be “all of the above.”

Dave Maass:

DHS entities published their AI use inventory—all AI they used or anticipate using— at the end of last year (under the Biden administration). The key thing to keep in mind with border surveillance is that these are things from administrations past, and we haven’t really seen any change in the trajectory; it just carries over from administration to administration. Land ports of entry and interior checkpoints have all sorts of surveillance tech. Customs and Border Protection (CBP) facilities also have a pretty comprehensive surveillance network at land crossings that will have cameras that can see for two miles, but also includes things like monitoring the facility and lines. This system is not just doing video, it is doing audio as well. That means when people are waiting in line, what they say may be recorded or listened to. Camera systems are also being combined with AI analytics.

Limitations on data-sharing

There are few protections for personal data in the United States. A lot of relevant data is held by state and local authorities, who may have to comply with state limitations on information and intelligence fusion centers, as well as some sanctuary city or state legislation. Unfortunately, many sanctuary cities do still share data with DHS, and few fusion centers are subject to meaningful oversight.

Third parties like social media platforms also hold a lot of data. The US does not have comprehensive privacy legislation, and under the legal precedent known as the “third-party doctrine,” law enforcement agencies can often access personal data without a judicial warrant.

Lawsuits like Centro de Trabajadores Unidos v. Bessent, challenging a data-sharing agreement between the Internal Revenue Service (IRS) and Immigration and Customs Enforcement (ICE), could provide insight into what exactly is happening, through discovery or court orders. In the meantime, here’s what experts had to say when I asked them what limits might exist.

Faiza Patel:

Rules for sharing will, of course, depend on the type of information at issue. For many types of data, such as that obtained through searches of devices at the border or collected by scraping social media, DHS rules for sharing are quite permissive. Indeed, DHS’s intelligence bases many of the intelligence bulletins it sends to thousands of police departments on social media posts.

Dave Maass:

It’s going to depend on the various states, their rules, and how cooperative the individual agencies are. For example, in some states, such as California, there are specific laws prohibiting agencies from accessing locally collected license plate reader data. Federal agencies can access CLETS ( the California Law Enforcement Telecommunications System) but have to say they won’t use it for immigration enforcement.

Technical limitations and threats

One unprecedented strategy from the Trump administration is the goal of accessing and combining vast amounts of data that were previously kept in separate systems. This month, ICE and IRS reportedly reached an agreement to share data for the purposes of deporting undocumented immigrants, and Elon Musk’s Department of Government Efficiency (DOGE) and the firm Palantir are building an API to access IRS records. How dangerous this will be depends on what they do with it. I asked experts how effective the Trump administration will be in using AI and other forms of surveillance.

Petra Molnar:

AI is just the latest obsession in the growing arsenal of tools that the Trump administration is touting as the panacea for border enforcement. However, we must also question the efficacy and accuracy of these tools, as often they just do not work or are not rolled out in the way that is presented. Artificial intelligence, in particular, is not actually very intelligent and can make mistakes or not work as intended.

However, this does not really matter because it is about the performance of surveillance as a deterrent to people on the move and marginalized communities.

Nathan Freitas:

Machine learning-powered face surveillance has become normalized, even in the lowest-stakes parts of our lives. Clearly, there is a high opportunity for false positives/negatives on many fronts, especially where the bias and error rate due to skin color cause an increased number of errors.

It should also be pointed out that we are moving from very specifically trained computer vision, algorithmic, machine learning models and classifiers, to more general “Large Vision Models” or “Multi-Modal Models”. These can be fed frames from video or still images, and asked to analyze them. (Author note: For example, conducting sentiment analysis based on visual cues.)

Raya Sharbain:

I definitely think tools like Cellebrite will be used and abused…Cellebrite has AI functionalities, so I wouldn’t be surprised if it was used to dream up some “evidence.” I’m familiar with tools like Dataminr, too, to ‘automate’ the administration’s work of scouring through social media platforms and publicly available data about people, and I do see it being added to border guards’ toolboxes.

Dave Maass:

New technology is being developed all the time. For example, a university in Texas has been hired to develop “eye detect,” a ”deceptive person” detector. This is, of course, incredibly suspicious in multiple ways. Once you add in all the different cultural and linguistic differences, or other factors such as people who have suffered trauma, it’s going to cause a lot of problems for a lot of people.

How can people lessen their risk?

Several security guides for crossing the border have recently been published, including one from The Intercept. In addition to understanding the technical risks, it’s also important to know your rights, especially the differences between the rights of citizens and non-citizens. Regardless of your status, there are some very specific steps that at-risk people can take.

Zahra Billoo:

For non-citizens, traveling with a burner phone could prompt further inquiry. For people with high-risk characteristics, what is actually more helpful than having a burner phone is that you consult with an attorney and have an attorney on standby.

For US citizens, the conversation about burner phones is frustrating because it would be better to resist, object, and understand the consequences of that. If we let this become normal for 3/12 years it will become normal. It is of utmost urgency that US citizens not be complicit in the attrition of their rights.

If you do not comply, the risk that you are taking is that your phone will be confiscated or you will face some type of delay and further questioning at the border…. at which point you should continue to assert your right to remain silent.” (Author’s note: US citizens do not have a general right to an attorney at the border, but do have the right to counsel if questions go beyond anything related to immigration.)

Nathan Freitas:

If I had one piece of advice: stop treating your phone like a sanctuary or diary. It is a communications device, and it is not loyal to you in any way. Otherwise:

• Have a circle of support group chat ready to assist you in border crossing or other tricky situations who can check-in before and after travel.

• If you feel at high risk, think about AirTags or other geotag devices that you can put in a shoe or other inconspicuous place and authorize them for tracking by a partner or loved one.

• The biggest areas of risk are photos, social media apps, and messaging apps, especially ones in which you are in large groups.

• The simplest thing is to remove sensitive apps and content from your phone, either at specific times or always. Only use social apps through a browser, and make sure to log out when encountering higher-risk situations like border inspections. Start using aliases online wherever you can, and avoid showing your face as much as possible.

• I do believe you have to provide your public social media account usernames. However, I don’t yet believe people are being forced to login into online accounts if their device is not already logged into them. This means, it can be helpful to set those accounts to private, if a search or analysis of those usernames is happening on another machine.

Raya Sharbain:

The advice here will always be: resist handing over your passcode or unlocking your device for police, no matter the context. In cases I’ve worked on, this usually hasn’t led to serious repercussions (as in people weren't detained as a result). In one case, an activist was denied entry into the country and their devices were seized, but were returned a week later. There was no indication that authorities succeeded in cracking the devices since the activist had secured them a strong alphanumeric passcode.

I see a few guides recommending burner phones or empty laptops, but I don't think that's practical advice for many reasons (expense, practicality, suspiciousness of empty or new devices). As for laptops, carrying an empty laptop across a border is not a terrible idea. You could argue with border guards that it's a work laptop.

Other essential practices:

• Turn off biometric unlock (face ID or fingerprint) – police can force you to use your face/fingerprint to unlock your device if you refuse to provide the passcode.

• Set a strong passcode (longer than six digits, relatively complex, preferably alphanumeric). Don't worry, this is temporary, just as you're crossing the border. You can re-enable biometric unlock later on.

• Right before crossing the border, turn off the device, and if seized, hand it over turned off. Don't turn it back on or unlock it (resist unlocking it).

• If you have some time to prepare, try doing some 'spring cleaning' on your device: go through your apps, messages, files, and photos, and consider what to keep and what to remove.

Other recommendations

I also spoke with some folks who have recently been through borders. They reminded me that sometimes what people in your social network have said is part of what you will be asked about, and that if you qualify, enrolling for the CBP’s Global Entry program can help concentrate risk at the time of application rather than leaving you stuck at a border.

Don’t let it become normal

Petra Molnar pointed out in her comments, "Technology is ultimately about power – and reinscribing the power differentials which are inherent in our world generally and in the immigration system specifically.”

If you are someone who holds more power, such as a US passport and the ability to pay for an attorney or find volunteer legal support, you have the ability to resist. The individuals cited in this article, and the organizations that they work for, are resources for keeping informed on what happens next. Zahra Billoo pointed out that there is a lot of organizing by grassroots organizations, including Know Your Rights trainings, legal support, and more. Take news articles with a grain of salt, search for that organizing, and prepare to protect yourself and your community now, rather than when you’re already being harassed by law enforcement.