Detecting Mal-Intent Before an Attack: FAST, SPOT, Quiet Skies, Secure Flight, and U.S. Traveler-Risk Analytics

Full report

On this page

Executive findings and analytical framework

The central finding of this investigation is that the United States has operated several very different kinds of aviation-security systems under the overlapping language of “risk,” “behavior,” “intent,” “vetting,” and “threat detection.” Treating them as one technological capability produces a misleading picture.

Secure Flight is principally an identity-resolution and screening-allocation system. It compares passenger biographical and itinerary data with watchlists and other high- and low-risk lists, then communicates a boarding or checkpoint-screening status. CBP’s Automated Targeting System, including its Unified Passenger interface, is a broader border-enforcement decision-support platform that combines passenger manifests, reservation records, prior encounters, watchlists, intelligence and rule-based indicators to prioritize travelers for review. Silent Partner and Quiet Skies used ATS-derived travel rules to identify non-watchlisted passengers for enhanced screening and, in Quiet Skies cases, sometimes Federal Air Marshal surveillance. SPOT and its successors attempted to infer risk from observable conduct. FAST went furthest conceptually: it was research intended to infer “malintent” from remotely sensed physiological and behavioral responses. Only the last two categories claimed, even indirectly, to detect an internal hostile state from a person’s current presentation.

That distinction produces five major conclusions.

First, FAST never became an operational passenger-screening system on the public record. It passed through laboratory research, controlled experiments, prototype development and limited field-assessment planning. DHS documents described goals, prototypes and successful research “events,” but those descriptions do not establish that FAST screened ordinary airline passengers for real-world security decisions. Its publicly documented lineage continued under names including “Passive Methods for Precision Behavioral Screening” and “Passively Activated Screening System,” but no authoritative public notice identifies an operational deployment or a precise termination date.

Second, SPOT was unquestionably operational, not merely research. TSA tested it beginning in October 2003, expanded it nationally from fiscal year 2007, employed thousands of dedicated Behavior Detection Officers, and spent roughly $900 million through the period GAO examined. Yet its most complete public outcome data showed arrests overwhelmingly associated with ordinary law-enforcement matters such as warrants, immigration violations, false documents, drugs and currency—not detection of terrorists preparing to attack aviation. GAO found that the scientific evidence did not establish that TSA’s indicators could reliably identify aviation threats.

Third, ending the standalone SPOT workforce did not end behavior detection. TSA eliminated the dedicated Behavior Detection Officer position in fiscal year 2017 and integrated former officers into the general screening workforce. As of November 30, 2021, approximately 3,730 officers remained qualified to conduct behavior-detection activities; TSA updated its behavior-detection operating procedures in May 2022. Referral rates fell sharply—from 13,187 referrals in 2016 to 506 in 2018—but the underlying function survived the organizational termination.

Fourth, Quiet Skies was an operational rule-based surveillance and screening program, not a sensor that detected intent. It identified some passengers who were not on a terrorist watchlist through travel-related rules, assigned enhanced screening, and from March 2018 sometimes caused Federal Air Marshals to follow and observe passengers on flights. TSA did not establish a comprehensive effectiveness measure, did not systematically connect marshal observations to rule validation, and could not use the absence of attacks as proof that Quiet Skies had prevented them. DHS announced the program’s termination on June 5, 2025, while stating that critical aviation-vetting functions would continue.

Fifth, Secure Flight and ATS remain active, and their core functions are more durable than individual program names. Secure Flight continues to operate continuously and normally receives passenger information within seventy-two hours of travel. ATS and Unified Passenger remain central CBP vetting platforms, integrated with the National Targeting Center and National Vetting Center. The end of Quiet Skies therefore did not eliminate watchlist matching, rules-based international passenger analysis, travel-pattern analysis, identity resolution, or risk-based screening allocation.

What counts as operational capability

This report uses a deliberately restrictive classification:

A promotional statement that a system “predicts hostile behavior,” a research accuracy figure from a balanced mock-crime experiment, or an allegation made in a congressional or executive-branch dispute is not treated as evidence that the capability worked operationally. Likewise, an arrest following a referral is not counted as a hostile-intent detection unless the public record connects the arrest to preparation for an aviation attack.

The four fundamentally different questions these systems answer

The programs are best understood as answering different questions:

Only the final layer approaches literal “pre-crime”: the attempted inference of future harmful intention without a known identity match, discovered weapon, completed offense or necessarily articulable external act.

Comparative lifecycle and present status

Comparative lifecycle timeline

Program-status table as of August 2, 2026

Program architecture, data inputs and decision consequences

FAST and Project Hostile Intent

DHS defined FAST’s target construct as “malintent”: a mental state involving an intention to cause harm to people or infrastructure. Research concepts involved remotely capturing cardiovascular and respiratory activity, facial temperature, eye movements, facial and body motion and vocal features. Some protocols contemplated presenting passive stimuli designed to activate mental representations associated with a planned harmful act, after which a classifier or sensor-fusion system would distinguish a “malintent” participant from a control.

The crucial limitation is that the labels were generated in experiments. Participants assigned to conceal a mock activity were not terrorists; the cost of detection was not imprisonment, death, mission failure or exposure of a real network; and experimenters generally knew the task structure and approximate class balance. A model that separates instructed “malintent” volunteers from controls in a controlled exercise has not thereby learned the physiological signature of an imminent terrorist attack.

Internal materials described a prototype capable of near-real-time off-body screening and concepts under which FAST would recommend accelerated or normal screening. Other documents discussed an airport-related pilot and kiosk testing. Those are meaningful stages of technology maturation, but the public record does not demonstrate routine screening of ordinary travelers or a binding acquisition decision. DHS’s fiscal year 2014 report calling the kiosk event a success is evidence that the event occurred, not that the system achieved transportable sensitivity, specificity or positive predictive value in the airline population.

FAST therefore belongs in the category laboratory research plus limited field-oriented prototype assessment. There is no public denominator showing how many genuine travelers it screened, how many it referred, how many referrals yielded weapons, or how many intended attacks it identified. No public confusion matrix demonstrates performance in a population with a realistic prevalence of hostile actors.

SPOT and integrated behavior detection

SPOT placed trained officers in or near airport screening environments. Officers observed travelers for behavioral indicators, sometimes engaged them in “casual conversation” or security-related questioning, assigned indicator points, and referred persons crossing a threshold for additional physical and baggage screening. Additional observations, document irregularities or discoveries could produce a referral to law enforcement.

Indicators were treated as deviations from an environmental baseline. Public examples have included changes in breathing, swallowing, excessive or unusual movements, apparent nervousness and responses during interaction. Many specific indicators and thresholds remain Sensitive Security Information. TSA’s operational claim was not necessarily that every displayed behavior meant terrorism; rather, clusters of indicators were said to justify additional screening because they could reflect stress, fear or deception associated with malintent. GAO documented an important conceptual inconsistency: at different times DHS characterized the indicators as detecting stress or fear while defending the program as one that identified malintent. These are not equivalent constructs.

When Congress reduced support for the standalone program, TSA did not abolish the observation function. It removed the dedicated job title, transferred thousands of personnel to general screening and retained qualified officers who could make behavior-based referrals. TSA officials told GAO that the activity became predominantly supportive of canine-enhanced screening. As of late 2021, qualification involved a forty-hour course, a knowledge test and twenty-four hours of on-the-job training.

This successor is narrower and produces far fewer referrals than the earlier SPOT structure. GAO found 13,187 referrals in 2016, or 182 per ten million screened passengers, compared with 506 in 2018, or six per ten million. Rates continued to fall through December 2021. This reduction is consistent with both a real contraction of the activity and a change in how referrals were recorded after dedicated officers disappeared.

Silent Partner and Quiet Skies

Silent Partner and Quiet Skies did not infer intent from a traveler’s face or physiology. TSA used CBP’s ATS infrastructure to evaluate international traveler records against intelligence-derived or experience-derived rules. Relevant data could include citizenship, itinerary, route, travel history, timing, reservation information, links or co-travel patterns, prior government encounters and law-enforcement or intelligence information. A match did not mean TSA had formally determined that the person was a terrorist or should be placed in the Terrorist Screening Database. It meant the person’s record matched a rule associated by the government with elevated risk.

Silent Partner primarily identified passengers inbound to the United States for enhanced checkpoint or gate screening. Quiet Skies was a subset that could place a passenger into a limited-duration status affecting later domestic or outbound travel. Secure Flight translated that status into a screening consequence, such as selectee treatment. In that sense, ATS supplied or supported the analytic match, while Secure Flight operationalized the checkpoint result.

From March 2018, Quiet Skies also became a resource-allocation input for FAMS. Investigative reporting based on leaked material described marshals recording whether selected travelers changed appearance, used telephones, went to lavatories, slept, stared, met other people or displayed other listed behaviors. The leaked duration rule was generally up to ninety days or three encounters. Those details are significant, but they should be described as leaked operational guidance corroborated by reporting—not as a complete official account of all versions of the program.

GAO found that FAMS after-action reports often said “nothing to report” or described an incident, but TSA Intelligence and Analysis did not regularly incorporate the reports into systematic rule evaluation. A program that generates covert observations yet does not consistently use those observations to estimate whether the triggering rule was valid has a broken feedback loop.

Secure Flight

Secure Flight requires covered carriers to transmit Secure Flight Passenger Data. Core elements include full name, date of birth, sex or gender information used for matching, itinerary and record-locator data; redress and Known Traveler Numbers are provided when applicable. Passport information and additional carrier data can be available in relevant international or risk-based contexts. Secure Flight compares records against the No Fly and Selectee components of the federal terrorism-screening system and other lists or risk categories authorized for aviation screening.

The principal consequences are:

• Cleared or ordinary screening;
• expedited screening for eligible low-risk populations;
• enhanced or selectee screening;
• inhibited boarding-pass issuance or denial of boarding for a confirmed No Fly determination; and
• resolution activity where the automated process produces a potential match.

A rules-based designation may result in more intensive screening, but TSA has represented that a person is not denied boarding solely because of general risk analysis; a prohibition ordinarily requires a No Fly or other legally operative restriction.

Secure Flight is therefore best described as an identity-matching and screening-orchestration system with added risk-based categorization. It identifies known or listed persons more directly than it predicts conduct. Its error modes include false identity matches, failure to match a listed person, incorrect list data, stale records and mistaken assignment to expanded or rules-based screening categories. GAO found TSA measured some potential and confirmed matches but did not measure every category of missed high-risk person or every false match to expanded and rules-based lists.

Federal Air Marshal surveillance

FAMS’s ordinary mission is not to classify every passenger. It deploys a scarce armed workforce to selected flights, using threat information, route risk, intelligence requirements and statutory priorities. Quiet Skies temporarily added a particular kind of trigger: the presence of a non-watchlisted passenger selected by a rules-based program.

This distinction matters when evaluating “detection.” An air marshal who observes a person and reports no suspicious activity has not validated the rule that selected the person. A marshal who notices an ordinary crime or disruptive act has not necessarily detected pre-attack terrorism. And an uneventful flight may mean the traveler was harmless, deterred, aware of surveillance, not yet operational, or selected by an overbroad rule. Without comparison groups and prespecified outcome definitions, none of those interpretations can be chosen empirically.

Quiet Skies’ termination ended that named special-mission trigger, according to DHS. It did not terminate FAMS, nor did it prevent FAMS from using watchlist information, intelligence leads, specific threats or broader risk-based flight-selection methods.

CBP Automated Targeting System and Unified Passenger

ATS-P has long been a web-based enforcement and decision-support tool. It applies rules and analyst review to Advance Passenger Information System data, Passenger Name Records and government holdings. APIS supplies manifest and identity data; PNR can contain itinerary, reservation dates, contact information, ticketing and payment information, travel-agency details, frequent-flyer fields, seat and baggage information, special-service requests and information linking travelers in the same reservation. ATS also connects to law-enforcement, immigration, watchlist and prior-encounter data.

Rules reflect intelligence, known smuggling or terrorist methods, suspicious trends and officer experience. ATS produces a relative risk assessment, alert, link or targeting result. Analysts at the National Targeting Center and officers at ports use that result to decide whether to conduct further vetting, issue instructions or recommendations, or refer a traveler for secondary inspection. GAO has emphasized that these tools augment rather than legally replace officer judgment.

Unified Passenger, or UPAX, is a technology refresh that replaces older ATS-P interface functionality. It does not represent abandonment of the ATS targeting model. CBP’s 2024–2028 information-technology strategy describes it as integrated with ATS and the National Vetting Center, including automation of traveler-vetting work.

ATS differs from FAST in three decisive ways. It does not need to claim that a person’s physiology discloses intent. Its features generally correspond to recorded facts, associations, government encounters or travel patterns. And its primary consequence is prioritization for human review or border inspection, not a scientific diagnosis of an internal mental state. Its legitimacy and accuracy nevertheless depend on rule quality, source-data quality, feedback loops, disparate effects and meaningful auditability.

Data-input and decision-consequence matrix

Scientific validity and quantitative performance

Construct validity: intent is not arousal

A valid hostile-intent detector would require a defensible chain of inference:

1. a person actually possesses a sufficiently defined harmful intention;
2. that intention generates a psychological state distinguishable from innocent states;
3. the state creates physiological or behavioral signals specific enough to hostile intent;
4. sensors or observers measure those signals reliably;
5. the classification rule generalizes across people, cultures, environments and attack types; and
6. the result remains useful at the extremely low base rate of real attackers.

Failure at any link undermines the final claim.

The National Academies’ polygraph review is directly relevant even though FAST was not a conventional polygraph. Heart rate, blood pressure, respiration and electrodermal changes reflect autonomic activation. They do not directly measure lying. Fear of false accusation, social anxiety, unfamiliar procedures, pain, medication, temperature, exertion, anger, trauma and disability can create similar activation. Conversely, practiced offenders, ideologically committed actors, people with atypical autonomic responses or subjects using countermeasures may show little detectable arousal. The Academies found that evidence for screening applications was particularly weak and cautioned against converting physiological reactions into direct conclusions about deception.

The distinction is sharper for hostile intent than for a specific lie. A lie-detection test can at least define a disputed proposition—“Did you place the object in the bag?” Hostile intent is temporally and normatively diffuse. Does an angry traveler who fantasizes about harm possess malintent? What about an undercover officer carrying out a red-team assignment, a protester intending civil disobedience, a passenger concealing an affair, or a person afraid because of immigration status? A system may detect concealment or stress without identifying the object of concealment, and may detect arousal without identifying an intention at all.

Human deception detection and behavioral cues

Bond and DePaulo’s meta-analysis synthesized 206 documents involving 24,483 judges. Average unaided accuracy was about 54 percent: judges correctly classified approximately 47 percent of lies and 61 percent of truths. That modest truth bias is not a basis for high-stakes airport triage, particularly where prevalence is far below the balanced or moderately balanced designs common in experiments.

The broader cue literature finds no equivalent of a behavioral “Pinocchio’s nose.” DePaulo and colleagues examined a very large set of verbal and nonverbal cues, most of which had small, inconsistent or context-dependent relationships with deception. Later work continues to find that very few bodily cues replicate consistently across studies. Emotional arousal, cognitive load and deliberate behavioral control can generate different—and sometimes opposing—presentations.

This does not mean behavior is always irrelevant. Specific conduct can rationally justify attention: attempting to evade a checkpoint, testing a locked door, abandoning a bag, surveilling security routines or providing materially inconsistent answers about a concrete fact. Those are observable actions with situational meaning. The scientific problem arises when generic signs such as gaze, swallowing, fidgeting, sweating, breathing changes or apparent nervousness are treated as transferable indicators of concealed criminal intent.

SPOT compounded that problem by asking officers to detect rare threats in a hurried, culturally diverse, high-stress environment. GAO’s review found substantial variation among airports and officers. Subjective baselines and indicator interpretations reduce inter-rater reliability: two trained officers may score the same traveler differently, while the same officer may score identical conduct differently depending on expectations, workload and context.

Ecological validity

Laboratory studies simplify reality in ways that generally make classification easier:

FAST’s claimed research success must therefore be discounted heavily when considering operational capability. The task was not simply to reproduce a laboratory metric at a checkpoint; it was to preserve discrimination under massive distribution shift and an extreme class imbalance.

SPOT faced a related ecological problem. Even if some behaviors correlate weakly with deception in structured interviews, passive observation of strangers walking through an airport is a different task. Cognitive-load interview methods sometimes improve information yield by asking unexpected questions or requiring detailed accounts, but that does not validate silent visual diagnosis of hostile intent.

Cultural variation, disability and observer bias

Behavioral meaning varies across cultures. Eye contact can signal confidence in one setting and disrespect in another. Speech latency may reflect deception, second-language processing or a culturally normative conversational rhythm. A 2023 cross-cultural study found that verbal patterns differed among South Asian bilinguals speaking Hindi or English and British monolinguals, illustrating the danger of assuming that a cue set developed largely from Western English-speaking populations is universal.

Disability creates comparable confounds. Autistic travelers may avert gaze, display repetitive movement, speak with unusual prosody, interpret questions literally, have difficulty with unstructured interaction or appear emotionally “flat.” Peer-reviewed studies have found that autistic presentations can cause observers to judge people as more deceptive or less credible despite the absence of deception. Airports themselves create sensory overload that can intensify atypical behavior.

Other confounds include Tourette syndrome, Parkinsonian movement, tremor, hearing impairment, post-traumatic stress, panic disorder, obsessive-compulsive behaviors, cognitive disability, speech disorders and side effects of medication. A procedure can be facially neutral while systematically referring certain populations more often because the proxy variables are not neutral in operation.

Observer bias can amplify those effects. An officer who expects a traveler’s clothing, language, religion, nationality or route to be risky may interpret ambiguous movement as suspicious. Once a conversation begins, confirmation bias can shape which inconsistencies are noticed and how behavior is scored. TSA’s anti-profiling policy is an important control, but GAO found in 2019 that supervisory mechanisms did not adequately direct officials to monitor unlawful profiling. TSA modified its checklists later that year; by 2022 the checklist expressly included race, religion, gender identity, disability and other protected characteristics.

A formal policy cannot by itself measure disparate effect. GAO found TSA did not have sufficiently comprehensive demographic data to determine whether screening practices referred particular passenger groups at disproportionate rates. Low complaint counts are also an unreliable proxy for low discrimination because travelers may not know the complaint process, may not know why they were selected, or may believe that complaining will accomplish nothing.

Inter-rater reliability and adversarial adaptation

For officer-based systems, reliability requires more than training completion. TSA would need blinded scoring studies in which multiple officers independently evaluate the same encounters, with prespecified measures such as Cohen’s kappa, intraclass correlation and test-retest stability. Public GAO work identified variation and subjectivity but does not disclose a robust nationwide reliability estimate demonstrating consistent application of the revised indicator set.

Adversaries also adapt. Once indicators become known through observation, leaks or training materials, an attacker can rehearse calm behavior, manufacture an innocent explanation, send a non-operational traveler to map screening or deliberately create misleading signals in other passengers. Physiology is not immune: breathing control, movement, medications, fatigue and habituation can alter signals. A classifier trained on naive participants may fail against strategic actors, the very population it is supposed to identify.

Rules-based travel targeting has its own adaptation problem. Networks can alter routes, use clean intermediaries, split reservations, avoid known transfer points or exploit people without prior derogatory records. The appropriate response is not necessarily to abandon rules, but to measure rule decay, false-positive rates and adversarial displacement rather than assuming that classified criteria remain effective indefinitely.

Machine learning does not cure weak constructs

FAST’s sensor fusion and modern traveler analytics are often discussed in machine-learning language, but a model cannot recover a target that has not been validly defined and labeled. If “malintent” labels are assigned by experimental role rather than independently verified mental state, the model may learn artifacts of the experiment: the wording of instructions, room location, participant demographics, sensor setup, time sequence or differences in how actors were coached.

Apparent high accuracy can also result from:

• evaluating repeated samples from the same participant in both training and test data;
• tuning features against a small test set;
• balancing guilty and innocent classes unrealistically;
• omitting inconclusive cases;
• failing to test at a new site;
• using proxy labels generated by earlier government decisions; or
• measuring discrimination within one dataset rather than calibration in the deployed population.

Recent cross-dataset research on automated deception detection continues to identify inconsistent cues, shortcut learning and near-chance transfer when models are evaluated outside familiar datasets. That evidence postdates FAST’s principal research period but reinforces the methodological concern: in-domain classification is not proof of a universal deception signal.

The base-rate problem

Suppose, purely for illustration, that one in one million screened travelers is preparing an aviation attack. Assume an extraordinarily strong detector with:

• 90 percent sensitivity; and
• 99 percent specificity.

For one million travelers, the system would detect about 0.9 genuine attackers but falsely flag about 10,000 innocent travelers. The positive predictive value would be roughly 0.009 percent—approximately one true case per 11,000 alerts.

Even increasing specificity to 99.9 percent would produce about 1,000 false alarms per million travelers, leaving fewer than one true case among roughly 1,001 alerts. The operational burden would include secondary searches, interviews, marshal deployments, database checks, missed flights, investigative review and diversion of officers from other threats.

This is why sensitivity figures from a balanced laboratory experiment are nearly meaningless without specificity and prevalence. Rare-event screening requires extraordinarily low false-positive rates, reliable ground truth and a decision consequence proportionate to uncertainty.

Quantitative outcomes and costs

What the SPOT numbers mean

The SPOT data provide the clearest public example of the denominator problem. Of approximately 61,000 referrals at the forty-nine airports in GAO’s fiscal years 2011–2012 dataset, approximately 365 ended in arrest. That can be presented as a 0.6 percent arrest yield among referrals, but it is not a terrorism-detection rate. Most arrests resulted only after additional database checks or searches and involved offenses not uniquely connected to the behavioral indicators.

Moreover, TSA’s validation study compared arrest outcomes among people who had different exposure to law enforcement. A referred passenger was more likely to have identification checked against warrant and immigration databases than an ordinary passenger. Increased arrest yield can therefore arise from increased inspection intensity, not from a valid ability to detect deception or terrorism. GAO criticized this unequal-exposure design.

The public data do not supply the missing categories:

Why the absence of attacks proves little

No successful attack during a program’s operation can be consistent with many competing explanations:

1. the program deterred or stopped attackers;
2. other layers—intelligence collection, overseas disruption, watchlists, hardened cockpit doors, passenger resistance, explosives screening or luck—prevented attacks;
3. no attacker attempted the type of operation the program was designed to detect;
4. adversaries selected other targets;
5. an attack was planned but abandoned for unrelated reasons;
6. the program was ineffective but the base rate was too low for failure to become visible.

Without a credible counterfactual, “no attack occurred” is an outcome of the entire security environment, not a program-specific performance measure. GAO’s Quiet Skies review illustrates the problem: TSA tracked rule matches and later watchlist additions but lacked a comprehensive evaluation design.

Privacy, constitutional law and civil rights

Fourth Amendment

Routine airport screening is treated in federal appellate law as an administrative search justified by the need to prevent weapons, explosives and hijacking. In United States v. Aukai, the Ninth Circuit held that consent is not the constitutional basis once a traveler enters the screening process; reasonableness instead depends on the regulatory purpose and the relationship between the intrusion and aviation security. Later Ninth Circuit cases continue to describe TSA screening as a Fourth Amendment search justified by that limited administrative purpose.

That doctrine does not give TSA unlimited criminal-investigative authority. The closer a search moves from detecting aviation threats toward a general search for drugs, currency, immigration violations or evidence of ordinary crime, the more serious the concern that the administrative rationale is being used as a pretext. Ninth Circuit precedent has specifically distinguished valid weapons-and-explosives screening from schemes designed in part to search for drugs or currency.

SPOT referrals therefore present a doctrinal tension. A behavior-based referral that triggers a proportionate explosives search at the checkpoint fits more comfortably within administrative-search doctrine. An escalating encounter whose practical purpose becomes discovering warrants, narcotics, immigration violations or unexplained cash is harder to justify solely by the original aviation-security rationale—although contraband discovered in a lawful security search is not necessarily suppressed.

Quiet Skies surveillance raised a different issue. Observation of a traveler’s movements, conversations and contacts in airports or aircraft cabins generally occurs in spaces where traditional doctrine recognizes limited expectations of privacy. Yet prolonged, technology-assisted or systematically aggregated monitoring can be more intrusive than an isolated observation. No Supreme Court decision has squarely resolved whether the full Quiet Skies pattern—secret algorithmic selection followed by repeated covert in-flight observation—constituted a search. The strongest constitutional question may therefore concern the combination of selection, duration, data retention and consequence rather than the observation of any one public act.

FAST would have raised a still more novel issue had it become operational. Remote sensing of cardiovascular, respiratory or thermal signals differs from ordinary visual observation because it seeks physiological information not necessarily consciously exposed to the public. Existing airport-search doctrine permits substantial screening for weapons, but a system designed to infer mental state from involuntary bodily signals could require a separate proportionality analysis. Because FAST did not publicly mature into operational deployment, courts have not supplied that analysis.

Fifth Amendment and due process

The Fifth Amendment’s Self-Incrimination Clause ordinarily protects compelled testimonial communication, not the collection of fingerprints, facial appearance, voice characteristics or other physical evidence. Passive physiological sensing would therefore not automatically constitute compelled testimony. The harder question would be whether the government used bodily measurements to extract the functional equivalent of a compelled statement about intention. Existing doctrine offers no clear answer because no operational FAST case reached the courts.

Due process is more concrete in watchlisting. Placement on the No Fly List can prevent travel, yet the underlying intelligence and criteria are often classified. In Kashem v. Barr, the Ninth Circuit rejected a facial vagueness challenge to the revised No Fly criteria and concluded that a reasonable-suspicion standard based on articulable intelligence constrained official discretion. The court evaluated procedures developed after earlier litigation, including notice that a person remained on the list and provision of an unclassified explanation when feasible.

In FBI v. Fikre, the Supreme Court unanimously held in 2024 that removing a plaintiff from the No Fly List and offering a limited assurance did not automatically moot his constitutional challenge. The decision did not decide whether the original listing was lawful. It did, however, prevent the government from defeating review merely by delisting a litigant without demonstrating that the challenged conduct could not reasonably recur.

Those cases involve formal watchlisting, which generally has clearer consequences and redress pathways than a Quiet Skies or ATS rule match. A person selected by a secret rule may experience repeated searches or surveillance without receiving notice of the reason, confirmation that a record exists, access to the rule or a meaningful opportunity to contest the inference. That “shadow classification” can be less severe than a No Fly prohibition but harder to challenge because the government may characterize each incident as an ordinary security measure.

Equal protection, religion and profiling

The federal government is constrained by the equal-protection component of the Fifth Amendment. A traveler alleging intentional racial, national-origin or religious discrimination would ordinarily need more than evidence that a neutral rule had unequal effects; proof of discriminatory purpose can be difficult when rules and intelligence sources are secret.

Behavioral screening is especially vulnerable to proxy discrimination. Race and religion may be formally prohibited factors, yet language, clothing, travel route, anxiety around police, eye-contact norms, family associations and names can influence either the rule or the human interpretation. The problem is not limited to purposeful prejudice. A poorly validated indicator can produce disparate referrals even when every officer acts in subjective good faith.

TSA policies prohibit profiling based on race, color, religion, national origin, ethnicity, sexual orientation, disability and other protected attributes. GAO nevertheless found that TSA’s earlier oversight mechanisms did not adequately direct supervisors to monitor profiling during behavior-detection activities. TSA revised checklists in October 2019, and later GAO work found supervisors were expected to observe whether referrals were made without regard to race, religion, gender identity, disability and related characteristics.

An earlier DHS Inspector General investigation did not corroborate a specific allegation of widespread racial profiling by behavior-detection officers at Boston Logan Airport. That limited finding should not be overstated. It does not establish that profiling never occurred nationally, nor does an allegation itself prove that it did. Civil-society reports documented complaints and statements from officers at several airports, but those records vary in corroboration and should be treated as evidence of risk and contested practice, not as a definitive nationwide prevalence estimate.

Religious profiling additionally implicates the First Amendment. Selection based on attendance at a mosque, religious association or protected expression would require a far stronger justification than selection based on individualized evidence of an operational plot. Fikre arose from allegations that included impermissible religious considerations, although the Supreme Court’s 2024 decision addressed mootness rather than adjudicating those allegations.

Disability law and accessibility

Section 504 of the Rehabilitation Act applies to TSA’s screening activities, and TSA provides accommodation procedures and passenger-support services. The legal risk is not confined to inaccessible equipment. A behavior system can discriminate when disability-associated conduct is treated as suspicious and produces more intensive searches, questioning or law-enforcement encounters.

GAO found that checkpoint technology and screening practices could more frequently refer transgender travelers, people wearing religious headwear, people with coarse hair, passengers using prostheses and persons with disabilities. For behavior detection, the comparable concern is that deviations from an assumed social baseline are not evenly distributed across the population.

Reasonable safeguards would include disability-aware indicator design, explicit prohibitions on scoring disability-related behavior, private accommodation, accessible notice, meaningful complaint retention, independent disparity audits and a mechanism to prevent repeated re-selection caused by the same misunderstood presentation.

Privacy Act, PIAs and systems of records

The Privacy Act regulates federal “systems of records” from which information is retrieved by personal identifier. Agencies must publish System of Records Notices describing categories of individuals, records, purposes, routine uses, retention and access procedures. Secure Flight and ATS operate through such notices and Privacy Impact Assessments. Yet law-enforcement and national-security exemptions can limit a traveler’s ability to learn what information exists or to correct it.

A PIA is a transparency and risk-management document. It is not proof that a system is scientifically valid, constitutionally reasonable or effective. A program can accurately describe its privacy controls while relying on an invalid predictor; conversely, a potentially useful program can still retain information too long or provide inadequate redress.

Secure Flight’s historically published retention framework allowed records with no watchlist match to be retained for up to seven days, potential matches for up to seven years and confirmed matches for ninety-nine years. These periods reflect different governmental interests but also produce different consequences from error: a mistaken confirmed match can become effectively lifelong.

ATS historically reduced its general retention period from forty years to fifteen years, with later records and modules governed by updated schedules and source-system limitations. PNR retention and access remain particularly sensitive because a reservation can disclose companions, contact information, travel agencies, payment arrangements, seat choices, special-service requests and long-term movement patterns.

Quiet Skies data crossed several systems and organizations: ATS supplied rule analysis, Secure Flight supplied screening status, FAMS created observation reports, and operational systems managed mission coverage. DHS OIG found that TSA’s use of some Quiet Skies passenger data was not entirely consistent with privacy guidance in the 2012 PIA. That is significant because privacy compliance depends not only on whether a collection was initially authorized, but whether later use, dissemination and retention remain within the stated purpose.

Watchlisting, notice and redress

DHS TRIP provides a redress process for travelers who experience repeated screening, mistaken identity or boarding problems. A redress number can help Secure Flight distinguish a traveler from a similarly named listed person. It does not necessarily reveal whether the person matched a classified ATS rule, appeared in intelligence reporting or was temporarily selected by a Quiet Skies-type process.

This creates three distinct redress problems:

The combination of secrecy and transience is especially difficult. A ninety-day Quiet Skies designation might expire before litigation, while retained records or analytic associations survive. Fikre limits one form of mootness gamesmanship in formal No Fly litigation, but it does not give every person selected for secondary screening a right to disclosure of classified criteria.

Aviation-security statutes and oversight

The Aviation and Transportation Security Act placed civil aviation security under TSA and authorized federal passenger screening. The Intelligence Reform and Terrorism Prevention Act of 2004 directed TSA to assume passenger watchlist matching, giving rise to Secure Flight’s central operational role. FAMS statutes authorize deployment on high-risk flights, while the TSA Modernization Act of 2018 imposed additional review and notification requirements concerning Silent Partner and Quiet Skies rules.

Statutory authorization to screen or target travelers does not answer whether a particular scientific inference is valid. Congress can authorize TSA to use behavior observation, but it cannot make swallowing, gaze or facial temperature a reliable indicator of terrorism by legislative declaration. This is why GAO’s methodological reviews were central: they tested whether the evidence justified the claimed screening value, not merely whether TSA possessed a broad security mission.

Classified and unavailable evidence

Several categories remain unavailable to the public:

• current Secure Flight matching thresholds and full error rates;
• most Silent Partner and Quiet Skies rules;
• ATS targeting rules, weights and rule-level yields;
• detailed FAMS flight-selection methods and coverage;
• current behavior-detection indicator thresholds;
• intelligence reporting used to nominate or associate travelers;
• complete terrorism-related referral outcomes;
• operational red-team results; and
• some cost allocations and contractor performance data.

Classification and Sensitive Security Information restrictions can be justified to prevent evasion. They also prevent independent researchers from reproducing agency claims. The appropriate conclusion is not that the programs secretly work or secretly fail. It is that public evidence cannot resolve the gap, and official assertions depending on unavailable evidence should be labeled unverified rather than accepted or rejected categorically.

Oversight findings, alternative explanations and continuity beneath renaming

Official defenses

Agencies have offered several recurring defenses.

TSA argues that layered security does not require each layer to identify a terrorist independently. A behavior referral may locate a weapon, expose fraudulent identity, disrupt reconnaissance or create uncertainty for an adversary. Rules-based targeting can prioritize limited officers even if a matched traveler is not formally suspected. Watchlist matching can prevent known threats from boarding while expedited screening permits resources to focus elsewhere. Secrecy protects criteria from countermeasures. These are plausible operational rationales.

TSA also revised SPOT indicators, altered training, reduced referrals and added anti-profiling oversight. Quiet Skies underwent working-group and stakeholder review, while ATS rules are reviewed using intelligence and inspection outcomes. The strongest official argument is therefore not that the programs literally read minds, but that imperfect indicators can still improve allocation at the margins when used with human judgment and additional screening.

That defense is strongest for Secure Flight, ATS and FAMS route selection, where the systems organize known information or allocate resources. It is weakest for generic behavioral or physiological claims, because marginal utility still requires that the indicators perform better than random, discretionary or alternative allocation after accounting for cost and bias.

Inspector-general, GAO and congressional critiques

GAO’s SPOT critique was scientific and programmatic. TSA had deployed nationally before establishing a sound effectiveness design. Its principal validation study contained methodological defects, indicator support was weak, and behavior-referral rates varied. GAO recommended constraining further funding until TSA could produce scientifically valid evidence.

GAO’s Silent Partner and Quiet Skies critique focused less on whether travel rules could ever be useful and more on absent evaluation. TSA counted rule matches, list size and some later watchlist additions, but did not establish a comprehensive metric. The retrospective analysis lacked a comparison group and covered Quiet Skies rather than the large majority of Silent Partner rules. FAMS reports were not consistently converted into validation evidence.

GAO’s Secure Flight critique was narrower. The core system operated, but TSA did not measure every kind of missed or falsely identified high-risk passenger. Much of the relevant accuracy evidence was withheld as Sensitive Security Information, limiting public evaluation.

DHS OIG’s Quiet Skies work identified governance, data-use and privacy deficiencies rather than proving that every selection was unlawful. Earlier OIG work on Secure Flight examined coordination and implementation, including the challenge of opaque match thresholds.

Congressional debate often followed partisan or episodic controversies. Assertions that a program targeted or protected particular political figures are relevant oversight allegations, but they are not self-proving. The June 2025 DHS announcement included claims of politicization and failure to stop any terrorist attack. The latter is an important official representation; the broader allegations require documentary or adjudicative corroboration before being treated as established fact.

Plausible alternative explanations for claimed successes

At least six alternative explanations can account for positive-looking program results without demonstrating hostile-intent detection.

Inspection-intensity explanation. Referred passengers receive more searches, questioning and database checks. Those procedures will discover more warrants, drugs, immigration violations and false documents even if the original behavioral indicator is unrelated to criminality. GAO identified this unequal-exposure problem in TSA’s SPOT validation.

General nervousness explanation. People carrying drugs, overstaying visas, traveling with false documents or fearing police may be anxious. A behavior system may therefore enrich for ordinary violations while having no capacity to distinguish an aviation attacker from other nervous travelers. Arrest yield can rise without validating the stated terrorism construct.

Shared-data or circularity explanation. A Quiet Skies rule and a later watchlist nomination may rely on overlapping intelligence, travel associations or source reporting. A person later appearing in the terrorism database is not an independent validation if both decisions emerged from the same underlying information. Apparent predictive success can be partly circular.

Broad-net explanation. A rule can identify some later-interesting people simply because it selects a large population. Without the denominator of all matched travelers and a comparison rate among unmatched travelers, the raw count of later watchlist additions is uninterpretable. This was a central defect in TSA’s Quiet Skies retrospective analysis.

Layered-security explanation. The absence of attacks may result from intelligence operations, overseas military or police action, physical screening, hardened aircraft, passenger intervention, cockpit security or a low number of attempted attacks—not the program under evaluation.

Behavioral displacement explanation. An adversary may shift routes, tactics or targets in response to the total security environment. That can be described as deterrence, but it does not show that a particular detector correctly identified anyone. Displacement could also impose costs elsewhere rather than reduce total risk.

Opportunity costs

Every false or low-value alert consumes a scarce resource. For SPOT, the opportunity cost included thousands of personnel, training, supervisory attention, secondary screening space and law-enforcement time. For Quiet Skies, it included air-marshal flight assignments that could have covered another route or threat. For ATS, overbroad rules can congest analyst queues and port inspections. For Secure Flight, weak matching can burden resolution staff and delay travelers.

Opportunity cost also has a security dimension. An officer focused on a nervous but harmless traveler is not observing another checkpoint area. A marshal following a rule-matched passenger may not be on a flight selected by a more current intelligence lead. Excessive false alarms can produce automation bias, alert fatigue and pressure to waive or shortcut review.

Civil-rights costs can become security costs as well. Communities that believe routine religious, ethnic or disability-associated behavior is treated as suspicious may become less willing to report threats or cooperate with investigators. Repeated unjustified screening can teach travelers to distrust redress processes and withhold context that would otherwise resolve an encounter.

Did termination or renaming end the underlying function?

The most important institutional lesson is that functions are more durable than labels. Data pipelines, legal authorities, trained personnel, interagency memoranda and user interfaces can be reassigned. A program may “end” for budgetary or political purposes while its narrower practices survive in ordinary operations. Conversely, the persistence of a data field or PIA after termination does not prove that the old operational mission is still being executed.

Quiet Skies is the clearest example. DHS’s termination announcement expressly said critical vetting would continue. Secure Flight still needs to compare travelers with watchlists and risk categories; CBP still analyzes PNR and APIS information; FAMS still allocates marshals using threat information. The termination removed a particular rules-and-surveillance package, not the federal traveler-vetting ecosystem.

What a scientifically defensible successor would require

A credible future program claiming to detect malicious intent would need, at minimum:

None of the publicly documented hostile-intent initiatives met that complete standard.

Final spectrum from ordinary screening to genuine “pre-crime”

The ranking below orders programs by how far they move from detecting a physical object or identifying a known person toward inferring future harmful conduct from indirect attributes. It is not a ranking of overall usefulness.

The ranking leads to a final distinction that should govern future oversight.

Secure Flight can be effective without predicting anyone’s intent if it accurately identifies listed persons and assigns the legally required screening status. ATS can provide value without proving that every rule-matched traveler is dangerous if its referrals improve inspection outcomes enough to justify their cost and disparate effects. FAMS can deter attacks without observing a crime on every covered flight if deployment is connected to defensible threat information.

SPOT and FAST require a more demanding showing because their rationale reaches inside the person. The government was not merely asking, “Is this the listed individual?” or “Does this itinerary warrant inspection?” It was asking whether breathing, swallowing, facial temperature, gaze, voice or demeanor revealed concealed hostile intention. The public scientific record does not support confidence in that inference at operational aviation base rates.

Accordingly, the evidence through August 2, 2026 supports the following bottom-line judgments:

The strongest public evidence favors systems that detect tangible threats, resolve identity, connect travelers to particularized intelligence, or transparently prioritize inspections. The weakest evidence concerns systems that infer deception or hostile intent from generic behavior and involuntary physiology. Renaming, reorganizing or terminating programs has repeatedly altered their visibility without necessarily eliminating their data flows or underlying screening functions. The central oversight task is therefore not merely to ask whether a named program still exists, but whether the same inputs, rules, personnel and consequences continue elsewhere—and whether they are validated against the specific claim the government makes for them.