The Sanctity of the Query: Reconceptualizing Search Privacy in the Age of Artificial Intelligence and Mass Surveillance

Full report

On this page

Introduction

In the digital epoch, the search engine has transcended its original function as a mere directory of information to become a foundational pillar of human cognition. Every day, billions of search queries are processed globally, capturing the unfiltered, real-time curiosities, anxieties, and intellectual explorations of the populace. These queries reveal questions individuals have not expressed publicly and beliefs they may not actually hold. Consequently, the search log represents a mirror of the human mind, reflecting a level of intimacy that arguably surpasses diaries, private correspondence, and library borrowing records. Despite the profound sensitivity of this data, the legal and regulatory frameworks governing search queries remain deeply fragmented. Operating under archaic legal doctrines formulated before the advent of the internet, modern search ecosystems routinely collect, retain, analyze, and monetize query data. The proliferation of artificial intelligence (AI) and machine learning (ML) has further exponentially escalated the sensitivity of these logs, enabling entities to infer highly sensitive, second-order characteristics from seemingly mundane behavioral records. Concurrently, law enforcement agencies have begun exploiting these centralized repositories through compelled disclosure and reverse keyword warrants, effectively transforming private intellectual inquiries into digital dragnets. This report conducts an exhaustive investigation into the epistemology of the search query, the architecture of modern search surveillance, and the intersection of First and Fourth Amendment jurisprudence. It examines the inadequacies of the third-party doctrine, evaluates recent judicial rulings and regulatory actions concerning electronic communications and data brokers, and proposes a comprehensive doctrine of Search Privacy. The analysis concludes by addressing whether lawful search queries should ever be utilized to assess an individual’s character, trustworthiness, politics, dangerousness, employability, insurability, or eligibility for services.

The Epistemology of the Search Query: Cognitive Offloading and the Extended Mind

To understand the necessity of protecting search queries, one must first understand their function in modern human psychology. The search bar is no longer just a tool; it is a repository for cognitive offloading, fundamentally shifting how the human brain processes and stores information. The Extended Mind Thesis, originally posited by philosophers Andy Clark and David Chalmers, argues that the mind is not confined within the biological brain but extends into the physical and digital artifacts used to process and store information1. The thesis relies on the "parity principle," famously illustrated by the thought experiment of Inga and Otto. Inga uses her biological memory to recall the location of a museum, while Otto, who suffers from memory degradation, relies on a notebook1. The philosophers argue that Otto's notebook plays the exact same cognitive role as Inga's neural pathways, and thus should be considered a constitutive part of his mind1. In the contemporary era, the search engine is the ultimate manifestation of Otto's notebook. When individuals delegate cognitive tasks to external aids, they alter their internal information processing requirements3. Research demonstrates that when individuals know information will be externally accessible, they exhibit "digital amnesia"—remembering the pathway to access the information rather than the information itself6. Because the search engine acts as a cognitive extension, the data it captures is fundamentally different from traditional communicative data. An email or text message is formulated for an external audience; it is filtered, curated, and subject to social norms. A search query, conversely, is a communication with the self. It represents the raw, unedited process of intellectual formulation. Recognizing the search engine as an extension of human cognition necessitates viewing search logs not as third-party business records, but as the modern equivalent of the brain's internal synaptic firings or a profoundly private diary of thought.

The Unspoken Revealed: AI, Machine Learning, and High-Level Inferences

Because search queries represent unfiltered cognitive processes, sequences of searches provide a comprehensive map of an individual's most private attributes. While a single search for a political figure or a medical symptom might indicate passing curiosity, longitudinal sequences of searches allow algorithms to infer highly accurate psychometric profiles. Artificial intelligence fundamentally increases the sensitivity of query logs because models are no longer limited to parsing the explicit text of a search; they infer higher-level characteristics by evaluating search frequency, timing, linguistic framing, and subsequent digital behavior7. The integration of machine learning into data analysis demonstrates that extraordinary inferences can be drawn across every facet of the human experience:

• Health and Medical Conditions: Landmark studies have demonstrated that anonymized web search logs can predict the future emergence of devastating diseases. Researchers from Microsoft demonstrated that by analyzing the escalation of searches from common symptoms to serious illnesses (a phenomenon linked to cyberchondria), they could predict the emergence of pancreatic adenocarcinoma months before a clinical diagnosis9. By identifying "experiential searchers" issuing first-person diagnostic queries, AI models accurately flag individuals for critical illnesses based entirely on their intellectual curiosity regarding their physical symptoms10.
• Mental State and Psychological Vulnerability: Machine learning models applied to search queries and social media texts can predict mental health states, including depression, stress, anxiety, bipolar disorder, and post-traumatic stress disorder13. Longitudinal analyses can track shifts in linguistic structure to identify individuals transitioning from general mental health inquiries to active suicidal ideation. For example, research indicates that users shifting toward suicidal ideation often exhibit a decrease in entity-based nouns and an increase in action-oriented verbs, coupled with a higher self-attentional focus using first-person singular pronouns15.
• Sexuality and Relationships: Search histories inherently capture unexpressed identities. AI classifiers can infer sensitive attributes, such as sexual orientation or relationship instability, even when these attributes are never explicitly disclosed in the query string16. Queries regarding local events, lifestyle choices, dating application support, or legal questions about divorce coalesce to form highly accurate predictive models regarding a user's intimate life.
• Finances and Employment: Users routinely search for solutions to immediate financial distress, querying payday loans, bankruptcy procedures, or unemployment benefits. When aggregated, these mundane behavioral records allow predictive models to accurately assess a user's financial precarity, employability, and socioeconomic status, often feeding into alternative credit scoring algorithms utilized by fintech companies18.
• Religion, Politics, and Philosophy: Intellectual exploration frequently involves researching diverse religious texts, political manifestos, or philosophical doctrines. An individual may query conservative fiscal policies, Marxist theory, or Islamic theology purely out of academic interest or private curiosity. However, AI inference engines routinely categorize users into rigid ideological buckets, stripping away the context of intellectual exploration and assigning definitive political or religious affiliations based on the frequency of such queries13.
• Criminal Law, Controversial Ideas, and Activism: Individuals use search engines to explore controversial philosophies, legal concepts, or the logistics of public protests. A crime novelist researching arson techniques, a law student querying explosive manufacturing statutes, or an activist researching geofence evasion techniques are all engaging in lawful intellectual exploration. However, automated safety classifiers and law enforcement algorithms lack the contextual nuance to distinguish academic curiosity from dangerousness, potentially flagging lawful private curiosity as criminal intent.
• Personal Fears: The search bar is a confessional for human anxiety. Queries regarding phobias, imposter syndrome, familial estrangement, or existential dread provide a roadmap to an individual's deepest psychological vulnerabilities.

The Architecture of Modern Search Systems: Processing, Retention, and Secondary Uses

The architecture of modern search systems is fundamentally designed to collect, retain, associate, analyze, and monetize user data. To adequately protect search privacy, one must distinguish between the varying methodologies of query processing and the functional pipelines that consume this data.

Search providers often publicly claim to "anonymize" search logs after a certain period to protect user privacy. However, these sanitization processes are frequently inadequate. For example, historical analyses of search retention policies have shown that the practice of obscuring the last octet of an IP address after 9 months, and deleting it entirely after 18 months, still leaves critical quasi-identifiers intact19. Because these bundled logs retain sufficient temporal and geographic information, they can be subjected to pseudonymization and subsequent de-anonymization attacks, rendering corporate promises of anonymization mathematically hollow19.

The Commercial Pipeline: Data Brokers, FCRA, and Alternative Data

The primary commercial utility of retaining exhaustive search histories extends far beyond simple ad targeting; it feeds a vast, unregulated data broker ecosystem. Data brokers systematically extract raw behavioral data—including search habits, geolocation, and digital transactions—stitching these points together to create complex inference profiles18. These profiles are subsequently sold to third parties, enabling discriminatory marketing, surveillance pricing, and the implementation of alternative credit scoring18. Fintech companies and modern lenders increasingly rely on "alternative data" to assess a consumer's creditworthiness. This paradigm shift means that a consumer's unexpressed online behavior, including their search history and browsing patterns, can directly dictate their access to loans, housing, and employment18. If an individual searches for information regarding debt consolidation or payday loans, data brokers can categorize them as financially distressed, resulting in targeted predatory lending or denied credit applications. The regulatory environment is only just beginning to recognize the severity of this crisis. The Consumer Financial Protection Bureau (CFPB) has initiated a rulemaking process to amend the Fair Credit Reporting Act (FCRA), seeking to broaden the definition of a "Consumer Reporting Agency" to include data brokers that sell personal identifiers and online behavioral data18. Under this proposed rule, the sale of "credit header" data (names, addresses, phone numbers, and social security numbers) and aggregated consumer reports used for eligibility decisions would be subject to strict regulatory oversight27. The CFPB's action effectively acknowledges that online behavioral data, derived largely from search and browsing histories, has become a de facto background check, necessitating transparency, fairness, and the right to dispute inaccuracies under the FCRA18.

The Legal Framework: Free Thought, Private Papers, and the Third-Party Doctrine

The legal distinction between a person's publicly expressed speech and their private intellectual investigation is central to the debate over search privacy. The United States Constitution provides structural protections for thought and papers, but these protections have been severely eroded by judicial doctrines adapted poorly to the digital age.

Intellectual Privacy and the First Amendment

The freedom of speech is hollow without a meaningful level of intellectual privacy. Legal scholars argue that intellectual privacy—the ability to develop ideas and beliefs away from the unwanted gaze or interference of others—is essential to a robust culture of free expression31. One must have the freedom to read, explore, and formulate ideas in private before one can have anything meaningful to say in public31. The Supreme Court has long recognized the constitutional necessity of anonymous reading and intellectual exploration. In Lamont v. Postmaster General, the Court struck down a Cold War-era law requiring citizens to affirmatively request the delivery of foreign communist propaganda, recognizing that government monitoring of reading habits creates an unconstitutional chilling effect on First Amendment rights34. This principle is fiercely protected in the physical realm. State laws, such as the Illinois Freedom to Read Act (Public Act 103-0100) and various Library Records Confidentiality Acts, strictly protect the privacy of patrons' reading habits, recognizing that the freedom to access information without surveillance is a fundamental human right36. However, a dangerous legal asymmetry exists in the modern era: while borrowing a book on a controversial topic from a physical library is heavily protected by statute and First Amendment precedent, typing that exact same topic into a search engine is routinely treated as a commercial transaction devoid of intellectual protection. The legal distinction between a person's expressed speech (which is protected) and their private intellectual investigation (which is commodified and surveilled) remains vastly inadequate.

The Fourth Amendment, Private Papers, and the Third-Party Doctrine

The Fourth Amendment guarantees the right of the people to be secure in their "persons, houses, papers, and effects." Historically, private diaries and papers were afforded near-absolute protection from government seizure31. However, the advent of the "third-party doctrine"—established in cases like United States v. Miller (bank records) and Smith v. Maryland (pen registers)—dictates that individuals lose their reasonable expectation of privacy in information voluntarily handed over to a third party39. Because users must transmit their search queries to corporate entities like Google or Microsoft to receive a result, law enforcement has historically argued that these queries are unprotected third-party business records, fully accessible without a probable cause warrant. This doctrine is facing profound judicial skepticism. In United States v. Jones, which involved the prolonged GPS tracking of a vehicle, Justice Sonia Sotomayor’s influential concurrence argued that the third-party doctrine is "ill suited to the digital age, in which people reveal a great deal of information about themselves to third parties in the course of carrying out mundane tasks"40. She explicitly noted that incredibly sensitive information—including trips to psychiatrists or abortion clinics—can be deduced from such data, and she questioned whether the warrantless disclosure of internet browsing records should be tolerated in a free society40. The Supreme Court further curtailed the third-party doctrine in Carpenter v. United States, ruling that the government generally requires a warrant to obtain historical Cell Site Location Information (CSLI) because such exhaustive digital records provide an intimate window into the whole of a person's physical movements43. The logic of Carpenter directly applies to search queries: much like location data, search histories are exhaustive, deeply revealing, and compiled automatically as an unavoidable byproduct of modern digital existence.

Scrupulous Exactitude and Expressive Materials

When a search warrant seeks materials protected by the First Amendment—such as books, political pamphlets, or correspondence—the Fourth Amendment's particularity requirement must be applied with "scrupulous exactitude." This doctrine was formalized in Stanford v. Texas, where the Supreme Court unanimously invalidated a sweeping general warrant authorizing the seizure of thousands of books and papers concerning the Communist Party of Texas from a private residence46. The Court noted the historical abuses of the Star Chamber and writs of assistance, declaring that when the "things" to be seized are books and the basis for their seizure is the ideas they contain, the warrant must be extraordinarily precise48. The Court reaffirmed this in Zurcher v. Stanford Daily, holding that while warrants can be issued against non-suspects, the preconditions must be applied with particular exactitude when First Amendment interests are endangered50. Because search queries are inherently expressive and intellectual, any government compulsion to access them must arguably meet this heightened standard of scrupulous exactitude. General demands for search logs represent the exact digital equivalent of the general warrants the Founders explicitly sought to abolish.

Secret Searches, Compelled Disclosure, and the Stored Communications Act

The statutory framework governing electronic privacy in the United States, primarily the Electronic Communications Privacy Act (ECPA) and its subset, the Stored Communications Act (SCA), is dangerously antiquated. Enacted in 1986, long before the advent of cloud computing and modern search engines, the SCA permits law enforcement to obtain vast quantities of digital information using a sliding scale of legal process based on how long the data has been stored51. For example, under the SCA, if an internet search or communication has been stored for more than 180 days, the government can compel its disclosure with a mere subpoena or court order based on "reasonable suspicion," entirely bypassing the Fourth Amendment's probable cause warrant requirement26. Furthermore, Section 2703(f) of the SCA allows any government agent to send a preservation request to an internet provider, compelling the company to copy and securely set aside an entire account's contents for up to 180 days25. This preservation ordinarily occurs without particularized suspicion, without a warrant, and in total secrecy from the user, triggering a Fourth Amendment seizure by stripping the user of control over their own data25. Hundreds of thousands of these secret government searches occur annually, fundamentally destabilizing democratic oversight and creating a chilling effect on freedom of expression25. The tension between compelled disclosure and intellectual privacy was prominently displayed in the 2006 case Gonzales v. Google. The Department of Justice issued a subpoena demanding a multi-stage random sample of one million URLs and a massive file containing the text of every search query entered into Google over a one-week period55. The government sought this data to test filtering software as part of its defense of the Child Online Protection Act (COPA)57. While other search providers complied, Google resisted the subpoena, citing the undue technological burden, the threat to its trade secrets, and the profound chilling effect the disclosure would have on user trust and privacy55. Although the district court ultimately compelled a heavily scaled-down production of URLs, it recognized the privacy concerns and shielded the actual search queries from disclosure61. Gonzales v. Google remains a landmark illustration of the immense value of aggregated query data and the perpetual threat of government overreach.

The Digital Dragnet: Reverse Keyword Warrants and Property-Based Approaches

The collision between massive corporate data retention and aggressive law enforcement tactics has\# The Sanctity of the Query: Reconceptualizing Search Privacy in the Age of Artificial Intelligence and Mass Surveillance

Introduction

In the contemporary digital epoch, the search engine has transcended its original architectural function as a mere directory of information to become a foundational pillar of human cognition. Every day, billions of search queries are processed globally, capturing the unfiltered, real-time curiosities, anxieties, and intellectual explorations of the populace. A search query fundamentally differs from other forms of digital communication. It reveals questions individuals have not expressed publicly and beliefs they may not actually hold. Consequently, the search log represents a high-fidelity mirror of the human mind, reflecting a level of intimacy that arguably surpasses diaries, private correspondence, library borrowing records, and medical dossiers. Despite the profound sensitivity of this data, the legal and regulatory frameworks governing search queries remain deeply fragmented and woefully inadequate. Operating under archaic legal doctrines formulated decades before the advent of ubiquitous broadband internet, modern search ecosystems routinely collect, retain, analyze, and monetize query data with minimal friction. The proliferation of artificial intelligence (AI) and machine learning (ML) has exponentially escalated the sensitivity of these logs. Advanced algorithmic models enable commercial entities and data brokers to infer highly sensitive, second-order characteristics—ranging from psychiatric deterioration to undisclosed sexual orientations—from seemingly mundane behavioral records. Concurrently, law enforcement agencies have begun exploiting these centralized corporate repositories through "reverse keyword warrants," effectively transforming private intellectual inquiries into indiscriminate digital dragnets. This report conducts an exhaustive, multidisciplinary investigation into the epistemology of the search query, the architecture of modern search surveillance, and the intersection of First and Fourth Amendment jurisprudence. It examines the inadequacies of the third-party doctrine, evaluates recent judicial rulings and regulatory actions, and ultimately develops a comprehensive doctrine of Search Privacy. The analysis concludes by addressing a fundamental societal question: whether lawful search queries should ever be utilized to assess an individual’s character, employability, or societal risk.

The Epistemology of the Search Query: Cognitive Offloading and the Extended Mind

To understand the necessity of protecting search queries with the highest degree of legal privilege, one must first comprehend their function in modern human psychology. The search bar is no longer simply an external tool; it is a repository for cognitive offloading, deeply integrated into the mechanics of human thought. The Extended Mind Thesis (EMT), originally posited in the philosophical literature by Andy Clark and David Chalmers, argues that the mind is not confined within the biological constraints of the brain1. Instead, mental states—including beliefs, desires, and memories—can be realized by physical processes and artifacts located outside the body1. When individuals delegate cognitive tasks to external aids, a process known as cognitive offloading, they fundamentally alter their internal information processing requirements2. Under the EMT, search engines serve as the ultimate external memory and associative processor. Empirical research demonstrates that when individuals know information will be externally accessible via a search engine, they exhibit "digital amnesia" or the "Google effect," meaning they remember the pathway to access the information rather than the information itself6. This shift systematically reconfigures the mind to act as a directory rather than a self-contained generator of thought6. Because the search engine acts as a cognitive extension, the data it captures is fundamentally different from traditional communicative data. An email, a published essay, or a text message is formulated for an external audience; it is filtered, curated, and subject to social norms. A search query, conversely, is a communication with the self. It represents the raw, unedited, and often chaotic process of intellectual formulation. The legal distinction between a person's expressed speech and their private intellectual investigation is central to this analysis. First Amendment scholar Neil Richards articulates that the protection of intellectual records is essential to the values of free thought and expression31. Intellectual privacy is defined as the ability to develop ideas and beliefs away from the unwanted gaze or interference of others31. Without a meaningful level of intellectual privacy, the freedom of speech is hollow; one must have the freedom to read, explore, and formulate ideas in private before one can have anything meaningful to say in public31. Recognizing the search engine as an extension of human cognition necessitates viewing search logs not as third-party business records, but as the modern equivalent of the brain's internal synaptic firings or a highly private intellectual diary.

Algorithmic Inferences: The Extraction of the Unspoken

Because search queries represent unfiltered cognitive processes, sequences of searches provide a comprehensive map of an individual's most private attributes. The integration of machine learning into data analysis has proven that extraordinary inferences can be drawn from search logs. Models are no longer limited to parsing the explicit text of a search; they infer higher-level characteristics by evaluating search frequency, timing, linguistic framing, and subsequent digital behavior. Consequently, data brokers and technology firms can extract a user's hidden reality, creating inference profiles that the users themselves may not fully comprehend. The scope of what can be potentially inferred from sequences of searches covers virtually every facet of the human condition: Health and Medical Conditions: Landmark studies have demonstrated that anonymized web search logs can predict the future emergence of devastating diseases months before a clinical diagnosis. Research conducted by Microsoft executives Eric Horvitz and Ryen White revealed that algorithms analyzing search logs could identify patterns of queries suggestive of an eventual diagnosis of pancreatic adenocarcinoma, a cancer with few early symptoms9. By analyzing the escalation of searches from common symptoms to serious illnesses—a phenomenon sometimes linked to "cyberchondria"—researchers can identify experiential searchers issuing first-person diagnostic queries11. Mental State and Psychological Vulnerability: Machine learning models applied to search queries and associated social media texts are increasingly utilized to predict mental health states, including depression, stress, anxiety, bipolar disorder, and post-traumatic stress disorder (PTSD)13. Furthermore, longitudinal analyses can track subtle shifts in linguistic structure, identifying individuals transitioning from general mental health inquiries to active suicidal ideation based on their digital footprint, verbs used, and decreasing cognitive coherence in their queries15. Religion, Politics, and Activism: While a user may never explicitly type "I am a member of a specific political party" or "I follow a specific religious denomination," their aggregate search behavior paints an unmistakable picture. Queries regarding local polling places, specific geopolitical news events, religious texts, or community organizing events allow algorithmic classifiers to accurately cluster users into highly specific demographic and psychographic categories16. The chilling effect on activism is severe if citizens fear their organizational queries are being monitored or algorithmically flagged. Sexuality and Relationships: Search engines frequently serve as the first confidant for individuals questioning their sexuality or navigating complex relationship dynamics. Sequences of searches regarding LGBTQ+ venues, relationship counseling, divorce attorneys, or reproductive healthcare can reveal an individual's sexual orientation or relational stability long before they choose to share this information with their spouse, family, or community16. Finances and Employment: Search histories heavily inform the creation of alternative credit scores. Fintech companies and data brokers utilize online behavior—including search histories indicating financial distress, job seeking, or gig-economy participation—to predict creditworthiness, employability, and loan risk18. An individual searching for "bankruptcy alternatives" or "unemployment benefits" may find themselves algorithmically categorized as a high-risk consumer, impacting their real-world economic mobility18. Criminal Law, Philosophy, Personal Fears, and Controversial Ideas: Individuals frequently use search engines to explore controversial philosophies, legal concepts, or criminal mechanics purely out of private curiosity or for academic research. A crime novelist researching arson techniques, a philosopher exploring the tenets of anarchism, or a citizen searching for information on the drug mifepristone all generate search logs that mimic illicit intent64. AI safety classifiers and risk models lack the deep contextual nuance to consistently distinguish between abstract intellectual exploration and genuine dangerousness, potentially flagging lawful private curiosity as a societal threat.

The Architecture of Modern Search Systems

To adequately protect search privacy, one must dissect the technical architecture through which modern search systems collect, retain, associate, analyze, personalize, monetize, secure, and disclose query information. The lifecycle of a query is complex and deeply integrated into the broader surveillance capitalist ecosystem. The following table distinguishes the various methodologies of query processing, association, and utilization employed by modern search platforms:

Search providers often publicly tout their efforts to "anonymize" search logs after a certain period to protect user privacy. However, independent analyses of these sanitization processes reveal critical vulnerabilities. For example, Google's historical retention policy involves obfuscating the last octet of an IP address after 9 months, and deleting the associated cookie after 18 months19. Because these data bundles retain the first three octets of the IP address, timestamp data, and the semantic content of the queries themselves, they maintain quasi-identifiers19. Attackers or data brokers can use these quasi-identifiers to link all searches made by a user and connect them to an arbitrary pseudonym, effectively executing a de-anonymization attack19.

AI Pretraining, RAG Architectures, and Privacy Leakage

Beyond traditional advertising, search logs are now the primary fuel for training Large Language Models (LLMs). The secondary use of search queries for AI development poses a permanent, non-revocable threat to intellectual privacy. When massive query logs are ingested into an AI's training corpus, the model memorizes patterns, syntax, and specific data points. Research into LLM pretraining and Retrieval-Augmented Generation (RAG) architectures demonstrates that embedding vectors and prompt caching can expose highly sensitive information22. If a user's highly specific, identity-revealing search query (e.g., a query containing a home address, a rare medical diagnosis, and a personal name) is ingested into a model's training data, sophisticated prompt extraction and model inversion attacks can force the LLM to regurgitate that private data verbatim21. Furthermore, side-channel timing attacks on prompt caching mechanisms in commercial LLM APIs can allow malicious actors to infer the contents of other users' recent queries, representing a profound privacy leakage23.

The Regulatory and Commercial Ecosystem: Data Brokers and Behavioral Profiling

The commercial pipeline relies heavily on the extraction of behavioral data to feed the secondary data broker market. Data brokers systematically extract raw behavioral data—including search habits, geolocation, and transaction histories—stitching these points together to create complex inference profiles such as "Propensity for Impulsive Decision-Making" or "Susceptible to Online Advertisements"18. These profiles are sold to third parties, enabling discriminatory marketing, surveillance pricing, and alternative credit scoring18. The regulatory environment is struggling to address this crisis. Recently, the Consumer Financial Protection Bureau (CFPB) initiated a highly anticipated rulemaking process to amend the Fair Credit Reporting Act (FCRA)27. The proposed rule seeks to bring data brokers within the scope of the FCRA by classifying them as "Consumer Reporting Agencies" if they assemble or evaluate consumer data for covered purposes28. Crucially, the CFPB proposal dictates that "credit header" data (names, addresses, phone numbers) and online behavioral data used for alternative credit scoring would be treated as "consumer reports"18. If finalized, this rule would require data brokers to adhere to strict FCRA requirements regarding accuracy, transparency, the right to dispute, and permissible purposes30. By attempting to regulate alternative data, the CFPB is effectively acknowledging that search histories and online behaviors have become de facto, unregulated background checks that heavily dictate a citizen's economic viability18.

Existing Protections and the Fourth Amendment Doctrine

The legal architecture protecting intellectual records in the United States is characterized by a stark asymmetry. While physical records of intellectual consumption are heavily protected, digital inquiries remain exceptionally vulnerable due to outdated judicial doctrines.

Library Records vs. Search Queries

The protection of physical reading habits is a well-established tenet of American privacy law. Statutes such as the Illinois Freedom to Read Act (Public Act 103-0100) and the Library Records Confidentiality Act strictly protect the privacy of patrons' reading habits and library borrowing records36. Law enforcement generally requires a highly specific court order to access a citizen's library checkout history. However, typing that exact same topic into a search engine is routinely treated as a commercial transaction devoid of intellectual protection.

The Stored Communications Act and Compelled Disclosure

The primary statutory framework governing government access to digital data is the Stored Communications Act (SCA), enacted in 1986 as part of the Electronic Communications Privacy Act (ECPA)51. The SCA is notoriously outdated and overly permissive. Under the SCA, law enforcement can routinely execute "secret government searches," compelling tech companies to produce user data while accompanied by injunctions forbidding the companies from notifying the targets51. Furthermore, under § 2703(f) of the SCA, any government agent can request the preservation of an internet account without particularized suspicion or probable cause25. The provider must copy the entire account and set it aside for up to 180 days, allowing the government time to build a case25. Transparency reports indicate that hundreds of thousands of accounts are preserved annually in this manner, operating as a massive, secret seizure of digital property25.

The Unraveling of the Third-Party Doctrine

The Fourth Amendment guarantees the right of the people to be secure in their "persons, houses, papers, and effects." Historically, private papers were afforded near-absolute protection. However, the advent of the "third-party doctrine"—established in 1970s cases like United States v. Miller (bank records) and Smith v. Maryland (telephone metadata)—dictates that individuals lose their reasonable expectation of privacy in information voluntarily handed over to a third party39. Because users transmit their search queries to corporate servers, law enforcement has historically argued that these queries are unprotected third-party business records. This doctrine is currently facing profound judicial skepticism. In United States v. Jones, which involved the prolonged GPS tracking of a vehicle, Justice Sonia Sotomayor’s influential concurrence argued that the third-party doctrine is "ill suited to the digital age, in which people reveal a great deal of information about themselves to third parties in the course of carrying out mundane tasks"40. She explicitly noted that incredibly sensitive information—including trips to psychiatrists, abortion clinics, and places of worship—can be deduced from such digital data, casting welcome doubt on the premise that transferring data to a tech company eliminates privacy rights40. The Supreme Court further curtailed the third-party doctrine in Carpenter v. United States, ruling that the government generally needs a warrant to obtain historical Cell Site Location Information (CSLI) because such exhaustive digital records provide an intimate, inescapable window into a person's life43. The structural logic of Carpenter applies directly to search queries: much like location data, search histories are exhaustive, deeply revealing, and compiled automatically as a byproduct of modern digital existence. Legal scholars increasingly advocate for a "property-based approach" to cloud data, arguing that users retain a possessory interest in their digital data (acting as a bailment), rendering search histories constitutionally protected "papers"72.

Search Warrants, Compelled Disclosure, and the Digital Dragnet

When a search warrant seeks materials protected by the First Amendment—such as books, political pamphlets, or correspondence—the Fourth Amendment's particularity requirement must be applied with "scrupulous exactitude."

Scrupulous Exactitude and Expressive Materials

This doctrine was formalized in Stanford v. Texas (1965), where the Supreme Court invalidated a general warrant authorizing the sweeping seizure of books and papers concerning the Communist Party46. The Court required the most "scrupulous exactitude" when the things to be seized are books and the basis for their seizure is the ideas they contain, noting the historical abuses of general warrants46. The Court reaffirmed this in Zurcher v. Stanford Daily, holding that the preconditions for a warrant must be applied with particular exactitude when First Amendment interests are endangered50. Because search queries are inherently expressive and intellectual, any government compulsion to access them must strictly meet this heightened standard. Early attempts to subpoena search logs in bulk faced resistance but ultimately demonstrated the government's appetite for this data. In Gonzales v. Google, Inc. (2006), the Department of Justice issued a subpoena demanding a multi-stage random sample of one million URLs and the text of millions of search queries over a one-week period to assist in defending the Child Online Protection Act (COPA)55. Google resisted, citing the technological burden and the protection of its trade secrets, but the court ordered a scaled-down production of URLs, highlighting the vulnerability of bulk query data to federal subpoenas55.

The Crisis of Reverse Keyword Warrants

The collision between massive corporate data retention and aggressive law enforcement tactics has birthed a highly controversial investigative tool: the reverse search warrant. Unlike traditional warrants that seek information regarding a known suspect, reverse warrants compel tech companies to search their entire databases of billions of users to identify anyone who engaged in specific conduct—such as searching for specific terms (keyword warrants) or being in a specific location (geofence warrants)72. A reverse keyword warrant demands that a search engine provider identify all users who searched for specific keywords or phrases within a defined timeframe26. The company executes a text-based query across its global database to locate responsive queries, returning identifying information (such as IP addresses and GAIA IDs) to the police39. These warrants operate as indiscriminate digital dragnets. They inherently lack Fourth Amendment particularity, as they do not name a specific person or place to be searched, instead casting a massive net over the intellectual activities of innocent civilians73.

Judicial Divergence and Legislative Pushback

The jurisprudence surrounding reverse warrants is deeply fractured. In 2023, the Colorado Supreme Court issued the nation's first appellate decision on a reverse keyword warrant in People v. Seymour. The case involved a tragic arson where police, lacking suspects, compelled Google to identify anyone who searched for the victim's address64. The Court made a groundbreaking ruling: it held that individuals possess a constitutionally protected privacy interest in their Google search history, even when revealed only in connection with an IP address, and that such searches implicate the First Amendment right to freedom of expression64. However, despite recognizing these profound constitutional rights and declaring the warrant defective for lacking individualized probable cause, the majority applied the "good-faith exception" to the exclusionary rule, allowing the evidence to be used65. As dissenting justices noted, this ruling offers a hollow victory; it validates the privacy interest but gives law enforcement a roadmap to rummage through the private search histories of a billion individuals without facing evidentiary suppression64. Conversely, federal courts analyzing similar reverse-location dragnets have taken a stricter approach. In 2024, the Fifth Circuit ruled in United States v. Smith that geofence warrants are categorically unconstitutional general warrants that violate the Fourth Amendment20. The structural logic of Smith applies equally, if not more forcefully, to reverse keyword warrants. While location data reveals where a body has been, query data reveals where a mind has traveled79. The use of reverse keyword warrants fundamentally chills protected First Amendment speech80. Recognizing this threat, state legislatures are moving to outlaw the practice. Lawmakers in New York, California, and Illinois have introduced legislation, such as the Reverse Location and Reverse Keyword Search Prohibition Act, aimed at banning law enforcement from utilizing these digital dragnets20. Furthermore, scholars advocate for amending the federal Stored Communications Act (SCA) to require "super-warrants"—akin to Title III wiretap authorizations—for any reverse search, demanding strict necessity and serious crime predication74.

A Proposed Doctrine of Search Privacy

The legal distinction between a person's expressed speech and their private intellectual investigation is no longer adequately maintained by existing frameworks. To protect the sanctity of the human mind in an era of pervasive surveillance and AI inference, the legal and technological communities must adopt a formal Doctrine of Search Privacy. This doctrine must supersede the third-party doctrine for intellectual queries and establish binding constraints on both the private and public sectors. The Doctrine of Search Privacy is constructed upon the following core pillars: 1\. Data Minimization and Identity Separation Search providers must architect their systems to guarantee that queries are decoupled from persistent user identities at the point of collection.

• Account-linked query retention should require explicit, affirmative, and granular opt-in consent from the user, unbundled from general terms of service.
• IP addresses and device fingerprints must be mathematically decoupled from the query string prior to storage, utilizing cryptographic hashing or differential privacy mechanisms that prevent retroactive de-anonymization via quasi-identifiers.

2\. Strict Retention Limits and User Deletion Search queries must be subject to aggressive, automated purging schedules.

• Retaining un-anonymized search logs indefinitely or for arbitrary periods (e.g., 18 months) poses unacceptable risks. Logs required for immediate system optimization, fraud prevention, or safety classifiers must be permanently deleted or mathematically anonymized within hours, not months.
• Users must possess an absolute, easily accessible right to delete their historical data, and this deletion must instantly propagate through all operational backups and downstream datasets.

3\. Prohibition on Secondary Use and AI Inference Training The exploitation of search queries for secondary commercial and generative purposes must be heavily restricted.

• Search providers must be prohibited from feeding account-linked or pseudonymous search histories into LLM pre-training pipelines or RAG databases. This is vital to prevent prompt-extraction attacks and embedding vector retrieval that result in irreversible privacy leakage21.

4\. Restrictions on Advertising, Profiling, and Automated Risk Categorization The commercial pipeline that treats intellectual inquiry as a tradable commodity must be severed.

• Data brokers must be strictly prohibited from aggregating search queries to build behavioral profiles for sale.
• Aligning with the CFPB's proposed FCRA rules, any entity utilizing search data to construct risk categories, alternative credit scores, insurability profiles, or employability metrics must be heavily regulated or banned entirely from using such data, due to the inherent lack of transparency and high risk of algorithmic bias18.

5\. Heightened Warrant Requirements and Dragnet Bans The Fourth Amendment must be interpreted to recognize a property and possessory privacy interest in digital search histories.

• Reverse Keyword Bans: Reverse keyword warrants must be categorically outlawed by statute and judicial precedent as unconstitutional general warrants. The government cannot search a billion minds to find one suspect.
• Individualized Scrupulous Exactitude: Warrants seeking the search history of a known, specific suspect must be subject to the heightened standard of "scrupulous exactitude" mandated by Stanford v. Texas46. Judges must require a demonstration that the targeted queries are directly instrumental to the crime, actively minimizing the incidental seizure of unrelated expressive and intellectual material.

Conclusion: Judging the Lawful Query

The final and most critical philosophical question regarding the architecture of modern surveillance is whether lawful search queries should ever be used to judge a person's character, trustworthiness, politics, dangerousness, employability, insurability, or eligibility for services. The answer must be an unequivocal and absolute no. To utilize lawful search queries as a metric for judging human character is to fundamentally misunderstand the nature of human cognition and to inflict devastating damage upon a free society. Because the search engine serves as a venue for cognitive offloading and intellectual exploration, it captures the mind in its most vulnerable and experimental state. A person who searches for symptoms of a chronic illness is not necessarily an uninsurable medical risk; they may be researching a character for a novel. A citizen who queries radical political ideologies is not necessarily a threat to the state; they may be an academic seeking to understand the mechanics of authoritarianism. An individual searching for bankruptcy laws is not necessarily an untrustworthy employee; they may be an attorney preparing for a complex litigation case. If society permits employers, insurers, data brokers, or the state to judge individuals based on their private intellectual investigations, it will precipitate a chilling effect of catastrophic proportions. The inevitable result is a phenomenon of cognitive foreclosure. Citizens will self-censor not just their speech, but their very thoughts, afraid to ask the digital ether questions that might result in a denied loan, a lost job opportunity, a falsely generated psychometric profile, or a law enforcement investigation. A democratic society relies on an informed, curious, and intellectually free populace. The legal distinction between a person's expressed speech and their private intellectual investigation must be fiercely guarded. Punishing individuals for the raw, unrefined inquiries they type into a search bar is tantamount to punishing them for their unformed thoughts. Therefore, search privacy is not merely a subset of data protection law; it is the fundamental prerequisite for the freedom of the human mind in the digital age.

Works cited

1. Right to mental integrity and neurotechnologies \- PMC \- NIH, https://pmc.ncbi.nlm.nih.gov/articles/PMC11503137/

2. An account of technologically extended semantic, episodic, and, https://www.cambridge.org/core/journals/memory-mind-and-media/article/an-overview-of-the-externalization-of-memory-an-account-of-technologically-extended-semantic-episodic-and-prospective-memory/214F056308C279323C93F9C7872FDEB4

3. AI Tools in Society: Impacts on Cognitive Offloading and the Future, https://dataethicsclub.com/write\_ups/2025-12-03\_writeup.html

4. Evan Risko on how technology affects the way we think, https://uwaterloo.ca/psychology/news/evan-risko-how-technology-affects-way-we-think

5. TheWeb‐Extended Mind \- ResearchGate, https://www.researchgate.net/publication/264370723\_TheWeb-Extended\_Mind

6. The extended hollowed mind: why foundational knowledge is, https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2025.1719019/full

7. Private traits and attributes are predictable from digital records of, https://pmc.ncbi.nlm.nih.gov/articles/PMC3625324/

8. Explorations of Intention and Utilization via Query Logs and User, https://erichorvitz.com/JAMIA\_2013\_diagnostic\_search.pdf

9. Analyzing Search Data Could Help Diagnose Serious Illness Faster, https://www.foxbusiness.com/markets/analyzing-search-data-could-help-diagnose-serious-illness-faster

10. Detecting Devastating Diseases in Search Logs \- Microsoft Research, https://www.microsoft.com/en-us/research/publication/detecting-devastating-diseases-in-search-logs/

11. Studies of the escalation of medical concerns in Web search, https://www.researchgate.net/publication/220515814\_Cyberchondria\_Studies\_of\_the\_escalation\_of\_medical\_concerns\_in\_Web\_search

12. Studies of the Escalation of Medical Concerns in Web Search, https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/TR-2008-178.pdf

13. An in-depth exploration of machine learning methods for mental, https://pmc.ncbi.nlm.nih.gov/articles/PMC12808420/

14. Mental Health Risk Detection From Social Media Text Data \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC13176780/

15. Discovering Shifts to Suicidal Ideation from Mental Health Content in, https://pmc.ncbi.nlm.nih.gov/articles/PMC5659860/

16. How Offensive AI Infers Sensitive User Information from Playlists, https://arxiv.org/html/2605.04724

17. The Curse of Dimensionality: De-identification Challenges in the, https://fpf.org/blog/the-curse-of-dimensionality-de-identification-challenges-in-the-sharing-of-highly-dimensional-datasets/

18. FCRA and Find Out: How Online Lenders Use Alternative Credit, https://greenlining.org/2025/fcra-and-find-out/

19. View of An Analysis of Google Logs Retention Policies, https://journalprivacyconfidentiality.org/index.php/jpc/article/view/593/576

20. Beyond Geofence Warrants: Chatrie, Reverse-Location Surveillance, https://www.criminallegalnews.org/news/2026/may/1/beyond-geofence-warrants-chatrie-reverse-location-surveillance-and-the-fourth-amendment-fight-over-digital-dragnets/

21. naacl2024.md \- GitHub, https://github.com/Doragd/Algorithm-Practice-in-Industry/blob/main/paperBotV2/conf\_summary/data/papers/naacl/naacl2024.md

22. Large Language Models with Temporal Reasoning for Longitudinal, https://pmc.ncbi.nlm.nih.gov/articles/PMC12702291/

23. Daily Papers \- Hugging Face, https://huggingface.co/papers?q=LLM%20API

24. Private Retrieval Augmented Generation via Random Projection, https://openreview.net/pdf?id=9swfLsr2f7

25. The Fourth Amendment Limits of Internet Content Preservation, https://scholarship.law.slu.edu/cgi/viewcontent.cgi?article=2264\&context=lj

26. Moderating Reverse Internet Keyword Warrants \- Scholarly Commons, https://scholarlycommons.law.northwestern.edu/cgi/viewcontent.cgi?article=1631\&context=nulr

27. CFPB Issues Proposed Rule to Cover Data Brokers Under the Fair, https://www.wsgr.com/en/insights/cfpb-issues-proposed-rule-to-cover-data-brokers-under-the-fair-credit-reporting-act.html

28. Consumer Financial Protection Bureau Releases Proposed Rule on, https://www.sidley.com/en/insights/newsupdates/2024/12/consumer-financial-protection-bureau-releases-proposed-rule-on-fair-credit-reporting-act

29. CFPB proposes FCRA rule with potential implications far beyond the, https://www.consumerfinancemonitor.com/2024/12/11/cfpb-proposes-fcra-rule-with-potential-implications-far-beyond-the-stated-focus-on-data-brokers/

30. What the Fair Credit Reporting Act Means for You | Mastercard US, https://www.mastercard.com/us/en/business/open-finance/a-summary-of-your-rights-under-the-fair-credit-reporting-act.html

31. Intellectual Privacy, https://openscholarship.wustl.edu/cgi/viewcontent.cgi?article=1508\&context=law\_scholarship

32. Intellectual Privacy \- ResearchGate, https://www.researchgate.net/publication/41664306\_Intellectual\_Privacy

33. Google has captured your mind \- Salon.com, https://www.salon.com/2015/02/26/google\_has\_captured\_your\_mind/

34. The Successes of the American Civil Liberties Union, https://www.aclu.org/successes-american-civil-liberties-union

35. Intellectual and Social Freedom (Chapter 20), https://www.cambridge.org/core/books/cambridge-handbook-of-surveillance-law/intellectual-and-social-freedom/FD946E7D9E761622FDF28EED7AD8EF09

36. Illinois Criminal & Traffic Law Manual \+ TacBook Digital Quick, https://blue360media.com/products/illinois-criminal-and-traffic-law-manual-with-tacbook-digital-quick-reference-guide

37. Illinois Strengthens Access to Information and Library Materials With, https://howweshareinformation.com/illinois-strengthens-access-to-information-and-library-materials-with-new-legislation/

38. Right to Read Act \- World Library Association, https://worldlibraryassociation.org/right-to-read-act/

39. Keyword Search Warrants and the Fourth Amendment, https://larc.cardozo.yu.edu/cgi/viewcontent.cgi?article=2181\&context=clr

40. Digest Case Commentary: United States v. Jones, http://jolt.law.harvard.edu/digest/digest-case-commentary-united-states-v-jones

41. United States v. Jones Decided? \- Berkeley Technology Law Journal, https://btlj.org/2012/03/united-states-v-jones-decided/

42. If These Walls Could Talk: The Smart Home and the Fourth, https://harvardlawreview.org/print/vol-130/if-these-walls-could-talk-the-smart-home-and-the-fourth-amendment-limits-of-the-third-party-doctrine/

43. Carpenter v. United States and the Emerging Expectations of, https://digitalcommons.wcl.american.edu/cgi/viewcontent.cgi?article=2111\&context=aulr

44. Carpenter v. United States: Big Data is Different \-, https://www.gwlr.org/carpenter-v-united-states-big-data-is-different/

45. How Sensitivity Can Categorize Third-Party Data After Carpenter, https://digitalcommons.law.uga.edu/cgi/viewcontent.cgi?article=1120\&context=glr

46. STANFORD v. TEXAS, 379 U.S. 476 (1965) \- FindLaw Caselaw, https://caselaw.findlaw.com/court/us-supreme-court/379/476.html

47. Stanford v. Texas | 379 U.S. 476 (1965) \- Justia Supreme Court, https://supreme.justia.com/cases/federal/us/379/476/

48. Stanford v. Texas \- Wikipedia, https://en.wikipedia.org/wiki/Stanford\_v.\_Texas

49. Stanford v. Texas, 379 U.S. 476 (1965) \- PastPaperHero, https://www.pastpaperhero.com/resources/stanford-v-texas-379-us-476-1965

50. James ZURCHER, etc., et al., Petitioners, v. The STANFORD DAILY, https://www.law.cornell.edu/supremecourt/text/436/547

51. Fixing the Stored Communications Act's Secret Search Problem, https://www.lawfaremedia.org/article/fixing-the-stored-communications-act-s-secret-search-problem

52. THE NEXT GENERATION PRIVACY ACT, https://fpf.org/kerr\_the-next-generation-privacy-act/

53. Secret Government Searches and Digital Civil Liberties, https://constitutioncenter.org/news-debate/special-projects/digital-privacy/secret-searches-and-digital-civil-liberties

54. Secret Searches and Digital Civil Liberties | Constitution Center, https://constitutioncenter.org/blog/secret-searches-and-digital-civil-liberties

55. GONZALES v. GOOGLE, INC. (N.D.Cal. 2006\) \- CaseMine, https://www.casemine.com/judgement/us/5914720cadd7b04934377cf8

56. The Information Privacy Law Project, https://openscholarship.wustl.edu/cgi/viewcontent.cgi?article=1525\&context=law\_scholarship

57. Effectiveness of Internet Content Filters, The, https://kb.osu.edu/server/api/core/bitstreams/37c4d11d-900b-5525-8fea-0f20e26cd157/content

58. PETER D. KEISLER \- Assistant Attorney General \- THEODORE HIRT, https://epic.org/wp-content/uploads/privacy/gmail/doj\_stark.pdf

59. Ruling in Google case a win for privacy? \- CNET, https://www.cnet.com/culture/ruling-in-google-case-a-win-for-privacy/

60. Gonzales v. Google Inc. \- Justia Dockets, https://docs.justia.com/cases/federal/district-courts/california/candce/5:2006mc80006/175448/12

61. what is google's reputation score? a method for modified self, https://scholarship.shu.edu/cgi/viewcontent.cgi?article=1923\&context=shlj

62. Cognitive offloading: The brain and its digital helpers \- ERGO Group, https://www.ergo.com/en/radar-magazine/digitalisation-and-technology/2025/cognitive-offloading-mental-performance-digital-helpers

63. Harnessing Internet Search Data as a Potential Tool for Medical, https://www.medrxiv.org/content/10.1101/2024.05.03.24305695v1.full.pdf

64. Colorado Supreme Court Upholds Keyword Search Warrant, https://www.eff.org/deeplinks/2023/10/colorado-supreme-court-upholds-keyword-search-warrant

65. Odd Colorado Ruling Upholds Internet Keyword Search Warrant, https://reason.com/2023/10/20/odd-colorado-ruling-upholds-internet-keyword-search-warrant/

66. Knowledge Integrity in Large Language Models: A State-of-The-Art, https://www.mdpi.com/2078-2489/16/12/1076

67. Found in Translation: A Generative Language Modeling Approach to, https://www.anuragkhandelwal.com/papers/fit.pdf

68. CFPB Proposes To Expand Fair Credit Reporting Act Implementing, https://www.arnoldporter.com/en/perspectives/advisories/2024/12/cfpb-proposes-to-expand-fcra-implementing-regulations

69. Comments to the CFPB on data brokers \- Upturn, https://www.upturn.org/work/comments-to-the-cfpb-on-data-brokers/

70. FCRA Rulemaking: A Path to Reining in Data Brokers \- Epic.org, https://epic.org/documents/fcra-rulemaking-a-path-to-reining-in-data-brokers/

71. Ban Book Bans, https://www.banbookbans.com/

72. A Property-Based Approach to Reverse Search Warrants, https://www.californialawreview.org/print/cloud-data

73. Fourth Amendment Constitutionality of Keyword Warrants, https://digitalcommons.law.uw.edu/cgi/viewcontent.cgi?article=5233\&context=wlr

74. Limiting Reverse Searches in the Stored Communications Act, https://www.lawfaremedia.org/article/limiting-reverse-searches-in-the-stored-communications-act

75. Case 1:23-mc-00067-MJS Document 1 Filed 07/07/23 Page 1 of 39, https://assets.bwbx.io/documents/users/iqjWHBFdfxIU/rLhYDa2V2Tn8/v0

76. People v. Seymour. \- Colorado Lawyer, https://cl.cobar.org/from-the-courts/people-v-seymour/

77. The Controversy Over Geofence Warrants and Privacy Rights, https://journals.law.umn.edu/mjlst/2025/01/20/caught-in-the-digital-dragnet-the-controversy-over-geofence-warrants-and-privacy-rights

78. United States v. Smith, No. 23-60321 (5th Cir. 2024\) \- Justia Law, https://law.justia.com/cases/federal/appellate-courts/ca5/23-60321/23-60321-2024-08-09.html

79. A Fourth Amendment of People and Places: Three Foundational, https://scholarship.law.marquette.edu/cgi/viewcontent.cgi?article=5597\&context=mulr

80. Death of Privacy in the Always On Future | Statute & Circuit \- Medium, https://medium.com/@delschlangen\_22694/the-death-of-privacy-in-the-always-on-future-1933824cd481

81. Reverse Search Warrants: Locating Google's Sensorvault Subjects, https://www.illiberalism.org/reverse-search-warrants-locating-googles-sensorvault-subjects-via-the-technological-illiberal-practice-of-surveillance-capitalism/