The black box sits beside the entrance to Woodland Glen IV, a residence hall on the University of Kentucky's campus in Lexington. Its job is simple to describe. You walk towards the door, it looks at you, and if it recognises the geometry of your face, the lock releases. No card, no phone, no fumbling. You are home.
The device is called Rock. It is made by Alcatraz AI, a company based in Cupertino, California, and this autumn the University of Kentucky began testing it at a handful of buildings, including two residence halls, as a way for students to get into the places where they sleep without tapping a key card. In mid-September local television stations covered the installation. The libertarian magazine Reason picked the story up nationally under the headline “Scan your face to enter your dorm?” The university, for its part, has been careful and consistent on one point: enrolling your face is voluntary. Your Wildcard, the university's student ID, still works. Your phone still works. Nobody has to do this.
That is true, and it is worth taking seriously. It is also where the interesting questions begin rather than end. For many of the eighteen-year-olds who arrived at Kentucky this autumn, this may be the first building they have ever been invited to open with their face. It is not a bank vault or an airport gate or a laboratory full of expensive reagents. It is the front door of their home: the one they walk through at two in the morning, after a bad exam, or carrying shopping bags with the key card buried somewhere beneath a textbook. What they learn at that door, about what is normal, about what is asked of them, about whether their face is something they give or something that is simply taken, may shape how they respond to every door that follows.
So here is the question the black box poses. When the student with the shopping walks up and the lock clicks open before they have put anything down, what have they actually experienced? Added safety? Added convenience? Or the first of many doors that will, one day, know who they are whether or not they asked?
What the University Actually Installed
Start with what is known, because the facts are more specific than the headlines.
According to the University of Kentucky Police Department's own information page, the biometric access pilot began at Woodland Glen IV. Residents were told to expect an enrolment email near the start of the semester, sent from an Alcatraz address, with personalised instructions. The page states that biometric data is collected solely to verify access to the building, that it is not used to track students' movements or behaviour beyond building entry, and that anyone the system fails to recognise can still get in with their Wildcard or mobile ID. WKYT, the Lexington CBS affiliate, reported on 19 September that scanners were also going in at Kirwan-Blanding, another residence complex, for a two-week trial. The Lexington Herald-Leader listed further planned sites, including parts of the medical campus, several biomedical research buildings and athletics practice spaces.
How the machine sees you was explained by the university's own police chief. Joe Monroe, who has led the UK Police Department since 2009, told LEX18 that the system “looks at your face, and it assigns different points using geometry, facial geometry. They call it facial vectors. Once that number is assigned, that is who you are. There's no pictures stored.” The university's privacy language, quoted by the Herald-Leader, describes the resulting “facial signature” as “a one-way encrypted mathematical representation of your facial features.” Alcatraz's chief executive, Tina D'Agostin, put it more bluntly to WKYT: “You're zeros and ones to us once you've been scanned.” Monroe has also conceded that a security brand named after America's most famous island prison “may not be the best name for the product, and we've pointed that out.”
Two details from the local coverage matter more than they first appear. The first is that this is not the university's first experiment with biometrics. According to the Herald-Leader, Rock is replacing iris-scanning equipment from a company called Eyelock that the university described as “non-functional.” Biometric entry at UK is an existing practice being upgraded and extended, not a novelty.
The second detail should stay with anyone thinking about consent. The Herald-Leader reported that enrolment would be required for people entering the most sensitive places on campus, including the police department and research facilities, and optional for less vulnerable places such as residence halls and most athletics facilities. LEX18 filmed the police evidence room, where Deputy Chief Nathan Brown explained that entry required both a credential and a face. “This is dual authentication because this is a very obvious restricted room,” he said. In other words, the University of Kentucky is not running a voluntary facial recognition programme. It is running one facial recognition system under two sets of rules. The Kentucky Kernel, the student newspaper, reported that the technology is also in other locations requiring two-factor authentication, such as research labs. For some people on campus, a face scan is already a condition of doing their job. For students, at their front doors, it is a choice. The technology is identical. Only the policy differs, and policy is the easiest part of any system to change.
A Company Built by a Face ID Veteran
Alcatraz AI is not a campus start-up that stumbled into student housing. It is an enterprise access control company with serious money and a clear ambition. It was founded in 2016 by Vince Gaydarzhiev, who, according to the company, led hardware prototyping for the iPad and iPhone at Apple during the development of Face ID. In a 2019 interview with Security Sales and Integration, Gaydarzhiev said the idea came from noticing that physical security, even at Apple's own campus, was still done “old school,” with badges and guards at almost every door.
In April 2026 the company announced a fifty million dollar Series B round, led by BlackPeak Capital, Cogito Capital and Taiwania Capital, bringing its total funding to more than one hundred million dollars. It claimed customers among the world's largest AI data centres, major US airports, energy companies, NFL teams, Fortune 100 firms and “major universities.” D'Agostin summarised the pitch in one line: “We are the Face ID of securing physical spaces.”
That framing tells you where the product is designed to end up. Face ID on a phone is not an optional extra for enthusiasts. It is the default way tens of millions of people unlock the most personal object they own. A company that describes itself as the Face ID of buildings is not aiming to sit politely beside the key card. It is aiming to make the key card feel as dated as typing a passcode. D'Agostin said as much to WKYT: “The badge, what most campuses are using today is a 30-year-old technology.”
Alcatraz's marketing to universities makes the same argument with a student in mind. A February 2026 guide on its website, pitching “privacy-first facial authentication” to higher education, opens with an observation about young people: “Students today unlock their phones with a glance and verify banking transactions with their faces. Yet on campus, they still fumble with physical keycards.” It promises “the frictionless experience your students expect.” To its credit, the same guide also sets out a strong consent standard, stating that students “must have the right to withdraw consent and delete their templates at any time.”
The guide is also candid about where else the technology could go. Beyond residence halls, labs and athletics, it suggests that “Libraries can automate entry for late-night study sessions,” and it mentions exam proctoring. There is no public evidence that the University of Kentucky plans to put face readers on its libraries or in its exam halls. But the people selling the system already imagine a campus where the face is the credential for much more than the dorm.
The Door That Watches Everyone Else
“Voluntary enrolment” describes the relationship between the system and the person who signs up. It says rather less about everyone else who walks through the door.
The main security argument for Rock is not really about the enrolled student. It is about tailgating: someone with no right to be in a building slipping in behind someone who does. D'Agostin told WKYT that the system lets security teams “detect when this incident occurs so they can react to it in real time.” The university has framed the housing tests in the same terms, as part of a broader effort to reduce unauthorised entry and strengthen residence hall safety, with the face option added as a hands-free extra rather than a replacement for existing credentials. A student interviewed by WKYT said they liked it “just because safety is the number one priority for the college.”
To detect a tailgater, though, the machine has to pay attention to people who have not enrolled. Alcatraz's product page is explicit. The reader “authenticates the enrolled person in under a second, then keeps monitoring the doorway as the door opens,” and if someone enters without authenticating, “a picture is captured at the door” and logged for investigation. The security operations centre, the page says, receives a snapshot of that person along with the time of the event and the device that triggered it. A 2021 announcement of the feature, reported by Security Systems News, described the same behaviour.
Neither the university nor local reporting has said whether tailgating snapshots are switched on at Woodland Glen IV or Kirwan-Blanding, and it would be wrong to assume that they are. But the design matters regardless of how any one site is configured. In a residence hall, the person walking in close behind a resident is very often not an intruder. It is a roommate who left their card upstairs, a friend coming to study, a parent carrying a mini-fridge. Some of those people will have chosen not to enrol. Some will never have been asked. A system whose core safety feature involves noticing, and potentially photographing, the people who did not authenticate is a system whose voluntariness applies to only one side of the door.
This is the structural problem that has dogged every face recognition product sold as a convenience. The person who opts in gets the benefit. The people around them bear some of the cost, and they are not the ones who were asked.
Voluntary Is a Setting, Not a Promise
Consent is often spoken about as though it were a fixed property of a technology. With access control, it is closer to a configuration option.
Alcatraz's public help centre describes three ways to enrol someone. Manual enrolment takes a single badge swipe and a few prompts on the Rock's screen. Web enrolment sends an invitation to register with your own phone or computer camera, which matches the emailed process the University of Kentucky describes. The third mode is auto-enrolment. When a Rock is configured this way, the help centre explains, “each time the user swipes their badge and enters through the door, the Rock is learning their face.” It needs four swipes, at least five minutes apart, from people who already have access to that door. In 2019 Security Sales and Integration described this self-enrolment as requiring no administration or employee action, “switching to face-only mode autonomously as soon as the employee is seen enough times.”
The direction of travel in the product is towards less friction at sign-up, not more. In May 2026 Alcatraz announced platform updates that, according to Biometric Update, let administrators apply access control rules to enrolment and added visual guidance to speed up biometric registration. D'Agostin said the changes were about making sign-up simpler and continuing to “automate operations and deliver a seamless authentication experience.” In September, at the GSX security show in Atlanta, the company went further. It named its software layers: RockOS, the firmware on each device that handles enrolment, liveness detection, authentication and tailgating detection, and Rock Cortex, a platform for verification, user and device management, integrations, monitoring, analytics and compliance. According to SecurityInfoWatch, a guided enrolment process that removes the need for users to move or tilt their heads was slated for general availability in October 2026, which Alcatraz says will make onboarding faster and cut failed enrolments. None of that is sinister. Making sign-up easy is what any product team would do. But ease is not the same thing as deliberation, and a choice that takes two seconds is rarely a choice anyone remembers making.
There is no evidence that the University of Kentucky uses auto-enrolment in its halls. Its published process is invitation-based and opt-in. The point is not that anyone is being enrolled without their knowledge. It is that the hardware on the wall can operate under a quite different consent model, and the difference is a setting chosen by an administrator. The student's choice is protected by a policy, not by the physics of the device.
Policies drift, and the most instructive recent example is an airport. The US Transportation Security Administration has long described its facial comparison checks as optional. In May 2025 a bipartisan group of senators, including Jeff Merkley, John Kennedy, Ed Markey and Roger Marshall, reintroduced legislation to rein the programme in. “Americans have the right to opt out of using TSA's facial recognition at the airport, and we need to protect that right,” Merkley said. Kennedy complained that the agency scans travellers “without even making it clear that they can opt out.” Two months later the Algorithmic Justice League, founded by the computer scientist Joy Buolamwini, published “Comply To Fly?”, drawing on travellers' accounts from 91 US airports. According to Biometric Update's summary, 99 per cent of respondents said officers had not told them verbally that they could opt out, many experienced the scans as mandatory, and some who declined reported hostility and extra scrutiny.
Nobody made the scans compulsory. They did not need to. When the scan is the fast lane, when the signs are small, when refusal is treated as an inconvenience, “optional” becomes something you have to insist on rather than something you are offered. That is how convenience quietly becomes expectation. It does not happen through a decree. It happens through queues.
The media scholars Nora Draper and Joseph Turow have a name for the feeling this produces. In a 2019 paper in New Media and Society, they described “digital resignation”: the condition in which people want to control what companies know about them but feel unable to, and give up. They argued that this resignation is not a flaw in consumers but something cultivated by routine corporate practices that make resistance tiring and pointless. A student who opts out in a hall where most neighbours have opted in will not be punished. They will simply be the one still tapping a card, still the exception. Most people do not choose to be the exception for long.
Home Is a Different Kind of Door
Privacy theory offers a useful way to explain why a residence hall feels different from a laboratory.
In a 2004 article in the Washington Law Review, the philosopher Helen Nissenbaum proposed that privacy is best understood as “contextual integrity.” Every setting carries norms about which information flows are appropriate within it, and the violation lies not in the information itself but in its movement out of the context where it belongs. In the police evidence room, the norm is scrutiny; anyone entering is expected to account for being there, and a face scan fits. A home is governed by the opposite norm. It is the one place you are supposed to be able to come and go without explaining yourself. That a student's home is owned by an institution, patrolled by its police and opened with its credential does not erase the norm. It makes it more fragile.
Then there is the question of who is being taught what. Most first-year undergraduates at Kentucky in 2026 were born around 2007 or 2008. Apple introduced Face ID with the iPhone X on 12 September 2017, when they were nine or ten. For this cohort, unlocking a device with a glance is a childhood habit, which is exactly what Alcatraz's marketing leans on. The dorm door extends that habit from the private object in your pocket to a shared building run by someone else. Face ID, by Apple's design, answers to you. The building answers to its owner.
What that habit looks like in practice surfaced a few weeks into term. The Kentucky Kernel spoke to Leah Henderson, a first-year student living in Woodland Glen IV, who said the university had not yet told her about the technology. She first assumed the cameras at the doors were recording students. Once she learned their purpose, she called the system innovative and convenient. “It's kind of cool. Maybe I don't have to use my phone all the time to get in,” she said. Purani Kannan, a sophomore and resident advisor at Kirwan-Blanding, told the paper that many students don't know much about the technology yet. Nothing in either account suggests anything improper; enrolment remains opt-in. But together they show how quickly an unexplained camera becomes a welcome convenience, and how little of that journey passes through anything resembling deliberate consent.
Mark Andrejevic and Neil Selwyn, both at Monash University, examined facial recognition in schools in a 2020 paper in Learning, Media and Technology. They warned that it risked altering schooling “along divisive, authoritarian and oppressive lines,” and worried about how such systems normalise surveillance for young people, teaching them to treat constant identification as the ordinary texture of institutional life. Their focus was compulsory schooling rather than universities, and Kentucky's pilot is opt-in. But the point travels: habits formed in institutions shape expectations outside them.
Not all research points one way. A 2022 study in Frontiers in Psychology surveyed 391 college students in mainland China and Taiwan about face recognition access control on their campuses. Students who found the systems useful and easy were more likely to use them, and that acceptance was associated with a stronger sense of school identity and belonging. The authors acknowledged “information leakage anxiety” but did not measure it. The finding is a useful corrective: for some students, a door that recognises them may feel like a place that knows them. That is precisely why normalisation matters. A technology that feels welcoming is easy to accept without examining what else comes with it.
What a Card Swipe Already Knows
It is tempting to treat the face as the moment surveillance begins. On most American campuses, the key card has been quietly generating data for years.
UKNow, the university's news service, described in August 2024 what its Wildcard Mobile ID could already do: residence hall access, payment at UK Dining, rides on LexTran buses, verification at athletic events, borrowing from UK Libraries and entry to recreational facilities. The same article said the university hoped to extend the mobile experience to the entire campus community, with faculty and staff to follow. Credentials on campus do not stay in one place; they spread to wherever it is convenient to use them. Each interaction is, potentially, a timestamped record of where a student was.
What institutions can do with such records is not hypothetical. In 2018 Sudha Ram, a professor of management information systems at the University of Arizona, drew attention for research using anonymised data from the university's CatCard ID system to predict which first-year students would drop out, based on how often they swiped into halls, libraries and recreation centres and where they spent money. “Many times, students who tend to drop out are people who have not established social relationships and who haven't established a regular routine on campus,” Ram told Arizona Public Media. Contemporary reports said the model identified roughly 85 to 90 per cent of students who went on to leave. The work was framed as supportive. It is also a textbook case of function creep: data collected to open doors and sell sandwiches becoming raw material for behavioural prediction. It happened with cards. Nothing about a face makes it less likely to happen with faces.
The architecture makes the connection concrete. According to Alcatraz, a Rock profile is linked to the user's badge ID in the existing access control system, so the face simply stands in for the card. Wherever the card is already a key, the face could, with the right hardware and a policy decision, become one too. Nobody has announced plans to put face readers in Kentucky's dining halls, libraries or gyms. But the path from one door to many is an installation job, not a reinvention.
Beyond campuses, the warnings are sharper. In late 2022 Madison Square Garden Entertainment used facial recognition at Radio City Music Hall to identify and eject Kelly Conlon, a lawyer chaperoning her daughter's Girl Scout troop, because her firm was in litigation against the company. Security technology became a tool for settling commercial scores. In Western Australia in 2021, police obtained data from SafeWA, a Covid-19 check-in app whose users had been told only contact tracers would see it. The government rushed through a law restricting its use. “Because it is so important to maintain community confidence in SafeWA, we've come down in favour of nobody,” said Attorney General John Quigley. “Not the CCC, not someone under FOI, not the police. Nobody sees this information.”
Note the order of events. The promise came first, its breach second, and the law protecting it only after the public found out. Anyone weighing a campus assurance that biometric data will not be used for tracking should ask what Western Australians eventually had to ask: what, other than the institution's goodwill, makes that true?
Kentucky's Law Stops at the Campus Gate
In Kentucky, the honest answer appears to be: not very much.
The Kentucky Consumer Data Protection Act, signed by Governor Andy Beshear in April 2024, took effect on 1 January 2026. It treats biometric data used for identification as sensitive and requires opt-in consent before a covered business processes it. That sounds like exactly the protection a student would want. But the Act states that its core provisions “shall not apply to any” city, state agency or political subdivision, nor to any “institution of higher education.” The University of Kentucky is a public university; on the plain text, the state's new privacy law does not appear to reach it. The Act also provides no private right of action, so even where it applies, only the attorney general can enforce it.
Illinois's Biometric Information Privacy Act of 2008, the strongest such law in the country, would not necessarily help either. BIPA requires written, informed consent before a private entity collects a face template and lets individuals sue, which is why it has produced some of the largest privacy settlements in history. But it excludes state and local government agencies, and private universities have slipped through too. In November 2022 a federal court dismissed a BIPA class action against DePaul University over a remote proctoring tool that captured students' facial data, ruling that because DePaul administered federal student loans it counted as an exempt financial institution.
Federal law offers a partial safeguard. Regulations under the Family Educational Rights and Privacy Act define a “biometric record” as a record of one or more measurable biological or behavioural characteristics that can be used for automated recognition of an individual, listing facial characteristics as an example, and treat it as personally identifiable information. But FERPA governs disclosure of education records. It does not decide whether an institution should build a face-matching system, and it says nothing about keeping enrolment optional.
Some policymakers have drawn firmer lines. In September 2023 New York's education commissioner, Betty Rosa, banned the purchase and use of facial recognition in the state's elementary and secondary schools, concluding after a state technology report that the risks outweighed the claimed benefits. That ban covers children, not adult university students. But it shows that a government can decide some settings are inappropriate for the technology, rather than leaving the question to procurement officers one pilot at a time.
In Kentucky, then, the durability of “voluntary” rests largely on institutional policy and on vendor commitments. The Herald-Leader reported that the company says it keeps no cloud biometric database and shares no data with third parties or government databases. Speaking to the Kentucky Kernel, D'Agostin was blunter still: “Basically, your face is a digital key, and when you come to the door, if that key matches, the door opens.” Of the data behind it, she said: “We don't have access and we don't store it, so we can't sell the data.” The Kernel reported that the data remains with the university, which manages it. Dani Jaffe, the university's associate director of strategic and crisis communications, said: “All biometric data is stored on secure servers and is never shared with third parties.” She added that the university “will continue to evaluate and implement tools” as security needs evolve. That clarifies where the promise lives. If the vendor holds nothing, the templates' fate rests on the university's own policy, and the state law that would bind a private company does not bind the university. Those are meaningful assurances. They are also promises made by the parties who benefit from the system's expansion, enforceable mainly by the parties who made them.
The Campaign That Already Won Once
None of this is new. Six years before Kentucky installed its scanners, the digital rights group Fight for the Future and the student organisation Students for Sensible Drug Policy set out to keep facial recognition off American campuses, publishing a scorecard of institutions and pushing resolutions through student governments.
Their best-known victory came at UCLA. In early 2020, as the university considered adding facial recognition to its security cameras, Fight for the Future ran more than 400 photographs of UCLA athletes and faculty through Amazon's face recognition software against a mugshot database. According to Inside Higher Ed, the test produced 58 false matches, most of them people of colour. UCLA dropped the plan. “UCLA will not pursue the use of facial recognition software technology,” Michael Beck, the administrative vice chancellor, said in a statement, concluding that “the potential benefits are limited and are vastly outweighed by the concerns of the campus community.” Evan Greer of Fight for the Future said the victory would “send a pretty strong message to any other administration who is considering doing this.” By then around 50 universities had publicly said they had no plans to use the technology.
Kentucky's pilot differs in important ways. UCLA's plan would have scanned everyone in view of campus cameras; Kentucky's is an opt-in door lock that verifies one enrolled person at a time. A mismatch at a dorm door means using your card, not being flagged as a suspect, and IEEE Spectrum has reported that best-case false negative rates are around two in 1,000. Michael Murray, a University of Kentucky law professor who studies artificial intelligence, told LEX18 that the university seemed to be “taking it quite seriously” and trying “to roll it out with a very low threshold of risk,” while cautioning that AI “is not cost-free, and it's not risk-free.”
Still, the old unease has not gone away. Local online discussion, summarised by the Lexington Times, was dominated by privacy worries, with some commenters calling the scanners the start of a surveillance network and others doubting that a face reader would stop anyone following a resident through an open door. Reason's Elizabeth Nolan Brown questioned whether the security case justified the system at all, and suggested institutions might be “dreaming bigger” than a limited dorm pilot. There is no sign yet of an organised Kentucky campaign on the UCLA model. But the arguments that won in 2020 are still available, and a pilot is the stage at which they are easiest to make.
Safety, Convenience, or the First of Many Doors
So what does the student carrying the shopping actually experience when the lock clicks open?
On the evidence available: mostly convenience, described as safety. The system is genuinely easier than hunting for a card, and nobody should sneer at that. Losing a Wildcard late at night is a real problem, and a door that knows you is a real comfort. The safety case is harder to pin down. Its main security function depends on the door paying attention to people who have not enrolled, in a building where the person behind you is usually a friend, and the university has not published data on how often tailgating causes harm in its halls.
What the student is less likely to notice is the thing that matters most over time. They are not only choosing between a card and a face. They are helping to establish a norm. Every enrolment makes the face reader more useful, the card tapper more unusual and the next expansion easier to justify. That is not a conspiracy. It is how defaults work.
It would be unfair to say the University of Kentucky has done anything wrong. On the public record, it has kept the dorm pilot optional, maintained alternatives, avoided storing photographs and explained the system openly. That is better practice than many face technology deployments. But it would be naïve to treat “voluntary” as a permanent feature rather than a present policy, in a state whose new privacy law excludes public universities, using hardware that is already mandatory elsewhere on the same campus.
If voluntariness is to last, it needs to be written down in forms that are hard to quietly revise. A university serious about the standard in Alcatraz's own guide could publish, before any expansion, answers to a short list of questions. Which features are switched on, including tailgating snapshots of people who have not enrolled? How long are templates and event logs kept? The police department's pilot page does not say, promising only that details on data use and retention will be shared during enrolment. Who can see them, and when would they go to police? What would have to happen, and who would have to agree, before face entry reached libraries, dining halls or exam rooms? And will a card or phone always remain a fully equal alternative, rather than the slow lane?
None of those questions requires opposing the technology. They require only that the choice offered to students today is still a choice for the students who arrive next year.
The first door a young person learns to open with their face is a lesson in what their face is for. It can teach them that their body is a password they control, lent to an institution for a specific purpose on specific terms. Or it can teach them that being recognised is simply the price of coming home. The black box beside Woodland Glen IV cannot tell the difference. The people who decide how it is configured can.
