The Architecture of Moral Inference: Simulating the Institutionalization of Sexual Risk in Algorithmic Governance

Introduction: The Extraordinary Diversity of Consensual Adult Sexuality

Human sexual behavior is characterized by an extraordinary degree of diversity, fluidity, and variance. When examined through the lenses of cultural anthropology and psychology, the boundaries of sexual normalcy are revealed not as fixed biological imperatives, but as shifting sociocultural constructions. Within the strict boundaries of adult consent, individuals engage in a vast and complex spectrum of practices, ideations, and expressions. Adults may legally consume lawful pornography, entertain deeply held taboo fantasies they explicitly and voluntarily choose never to enact, and participate in unconventional but entirely consensual relationships. They continually seek sexual-health information, utilize sophisticated dating services to navigate complex social ecosystems, and engage in consensual kink and power-exchange dynamics. Some adults frequently change partners, while others choose strict celibacy. Many question their sexual orientation, explore the boundaries of their gender and sexual identity, read explicit erotic fiction, or discuss highly stigmatized fantasies in trusted, private digital spaces. Above all, many adults choose to keep their sexual interests strictly private, viewing their desires as an intimate extension of their innermost psychological lives.
Because of this profound variance, behaviors that are considered entirely normal, healthy, or culturally celebrated within certain populations are frequently stigmatized, pathologized, or even rendered illegal in others. The parameters of this analysis are strictly confined to the conduct of consenting adults. The sexual exploitation and abuse of minors, or any conduct involving coercion or nonconsenting individuals, are structurally and legally distinct categories of harm. Such actions fall entirely outside the parameters of normal consensual behavior and are categorized as absolute violations of both law and human rights.
However, as the digital infrastructure mediating modern life becomes increasingly governed by artificial intelligence and algorithmic moderation, a profound civil liberties challenge emerges. The convergence of expansive age-assurance mandates, rigorous consent verification protocols, heightened platform safety duties, aggressive sexual-content classification systems, and sweeping identity verification regimes threatens to transform the private sphere of adult sexuality into a legible, trackable, and ultimately evaluable dataset. This report conducts a rigorous future simulation, investigating how machine systems, originally deployed for the legitimate purpose of regulatory compliance and child safety, could inexorably begin treating ordinary, consensual adult sexuality as an institutional risk variable.

The Current Legal and Technological Baseline

To understand the trajectory of future algorithmic governance, it is necessary to establish the current legal and technological baseline, distinguishing the reality of 2026 from the subsequent simulation. Over the past several years, a wave of legislation across global and state jurisdictions has fundamentally altered the architecture of digital anonymity and content access, primarily driven by a legislative desire to protect minors from explicit material.
The legal watershed occurred with the U.S. Supreme Court's 2025 decision in Free Speech Coalition, Inc. v. Paxton1. The Court upheld Texas House Bill 1181, which mandated age verification for websites where more than one-third of the content is classified as "sexual material harmful to minors"2. Departing from the strict scrutiny frameworks historically applied to adult access to protected speech—such as the landmark 1997 Reno v. ACLU decision—the Court applied intermediate scrutiny3. Justice Clarence Thomas, writing for the 6-3 majority, ruled that the burden on adults was merely "incidental" to the state's compelling interest in shielding children, asserting that no person has a First Amendment right to access obscene-to-minors content without submitting proof of age2.
In a vigorous dissent, Justice Elena Kagan highlighted the severe privacy risks, arguing that age verification forces users to turn over highly sensitive information about their viewing habits regarding socially repulsive speech to unknown website operators who might sell, leak, or be subpoenaed for the data3. Nevertheless, the ruling catalyzed a cascade of state-level age verification laws across the United States. By 2026, over half of U.S. states—including Louisiana, Utah, Virginia, Arkansas, Indiana, Idaho, Florida, and Georgia—had enacted legislation requiring digital service providers to utilize government-issued identification or commercial data systems to verify user age5. Internationally, frameworks like the United Kingdom's Online Safety Act, France's social media bans for minors, and the European Union's Digital Services Act (DSA) have similarly mandated age assurance and platform safety duties8.
While the stated intent of these laws is the protection of minors, the technological implementation necessitates a massive expansion of surveillance infrastructure. To verify that a user is an adult, platforms must often rely on third-party identity brokers, biometric facial age estimation, document-based verification, or credit-card checks8. Proponents of these laws frequently argue that privacy-preserving architectures, such as Zero-Knowledge Proofs (ZKPs) and W3C Verifiable Credentials, can solve the privacy dilemma9. The W3C Verifiable Credentials data model supports ZKPs to allow "selective disclosure" and predicate proofs—for example, mathematically proving a user is over 18 without revealing their exact date of birth or name to the verifier12.
However, as digital rights groups like the Electronic Frontier Foundation point out, cryptographic solutions do not eliminate the upstream identity verification problem9. The verifiable credentials must still be issued by a central authority or identity broker that knows the user's true identity9. Furthermore, the verification transaction itself generates metadata that can link an identity to a specific site or intent10. Consequently, the requirement that adults identify themselves to access protected speech functions as a de facto surveillance system, effectively terminating the era of anonymous browsing for sensitive content13. The law normalizes this surveillance through what legal scholars term "privacy nicks"—the gradual, systematic conditioning of the public to accept smaller, frequent, and mundane privacy diminutions that cumulatively acclimate society to being watched in increasingly intimate ways16.

Simulating Mission Creep: The Seven Stages of Algorithmic Overreach

With the baseline of mandatory identity verification and age assurance established, the simulation projects these mechanisms into the near future. Mission creep, alongside its corollaries of function creep and surveillance creep, occurs when sociotechnical systems designed for a specific, narrow purpose gradually expand their operational scope to encompass new targets and objectives, often leading to abuses of power or incompatible secondary uses17. In the context of algorithmic governance, data collection justified by child safety or platform security is highly susceptible to function creep, evolving to analyze peripheral societal functions9.
The following seven-stage simulation models how machine systems, driven by market incentives and risk-aversion, will logically evolve to treat adult sexuality as a comprehensive, institutional risk variable.

In Stage 1, the machine merely asks a binary question: Are you an adult? This necessitates the establishment of a verified digital identity9. In Stage 2, the machine asks: Is this content sexual, and does it require gating? This requires the deployment of advanced natural language processing and computer vision to categorize the entirety of human sexual expression20.
The critical, irrevocable pivot occurs in Stage 3. Machine learning algorithms are fundamentally designed to optimize engagement, reduce latency, and predict user behavior. If a system knows that a specific user consistently accesses alternative sexual content, it can infer a latent interest. By observing engagement with specific erotic fiction, searches for niche sexual health queries, and interactions with specific dating profiles, the machine generates a multidimensional vector representation of the user's sexual psyche. The user never needs to declare their interests; the machine infers them through behavioral exhaust.
By Stage 4, the retention of this highly sensitive data is paradoxically justified as a legal necessity. Platforms, terrified of catastrophic liability under expanded safety duties, maintain detailed logs to prove to regulators and auditors that they are accurately age-gating the correct populations. This creates a honeypot of unparalleled sensitivity, merging verified physical identities with latent sexual desires10.
In Stage 5, the logic of algorithmic risk management supersedes content moderation. Machine learning systems are inherently associative; they seek patterns across massive datasets. Suppose the machine notices a weak statistical correlation between users who consume aggressive, dominant-submissive pornography and users who receive complaints for using abrasive language in non-sexual digital forums. The algorithm, devoid of contextual or semantic understanding, creates a mathematical link. The sexual interest is now tagged in the system as a "toxicity predictor."
In Stage 6, the financial imperatives of surveillance capitalism drive the syndication of this data. The boundaries between digital platforms, data brokers, and institutional evaluators dissolve. The risk profile is commodified. Finally, in Stage 7, the loop is closed, and the simulation becomes a socioeconomic reality. A citizen applies for a corporate leadership position. The employer's automated background-screening software queries a federated data broker. The applicant's hidden "sexual risk vector"—derived entirely from their private consumption of lawful, consensual media—causes the algorithm to flag them as a "high-risk candidate for workplace instability." The application is summarily rejected without human review. The individual's socioeconomic fate has been decided by their private sexual desires.

The Transmutation of Statistical Association into Moral Judgment

The most dangerous systemic failure in Stage 5 of the simulation is the algorithmic transmutation of statistical association into moral and administrative judgment. Algorithms do not understand causation, nuance, or human psychology; they only understand mathematical correlation. When a machine analyzes millions of data points, it will inevitably find spurious or weak correlations between specific private sexual behaviors and negative external outcomes, such as divorce, workplace complaints, financial instability, mental distress, or criminality.
The leap from statistical correlation to adverse administrative decision-making relies on a profound epistemological flaw regarding actuarial risk. This flaw is heavily documented in the literature surrounding clinical and actuarial risk assessment tools, such as the Static-99 and Static-99R, which are used within the criminal justice system to assess the recidivism risk of individuals convicted of sexual offenses21. Actuarial instruments calculate a group-level probability based on static historical factors, assigning individuals to nominal risk categories22. However, applying a group base rate to an individual to determine their specific future conduct is highly unstable and mathematically fallacious, often generating devastating false positives24. Critics point out that the Area Under the Curve (AUC) metrics used to validate these models merely indicate how well the tool separates groups of recidivists from non-recidivists, not how accurately it predicts the risk of a specific individual, and that these models are often highly unstable across different sample populations24.
When algorithmic systems scale this flawed actuarial logic outside of the criminal justice system and apply it to the general, non-criminal population, the danger multiplies exponentially. The machine constructs a profile based on a person's digital footprints. If the machine's training data indicates that a fraction of a percent of users who search for a specific taboo fantasy also exhibit high rates of financial delinquency, the algorithm integrates that fantasy into a credit-risk model.
This represents the birth of algorithmic "actuarial justice" replacing individualized evaluation and due process26. The algorithm treats the individual not as a moral agent capable of choice, compartmentalization, and self-control, but as a bundle of statistical probabilities. The association itself becomes a moral judgment. The machine essentially declares that because people who share a user's private desires have statistically exhibited higher rates of divorce or workplace complaints, the user is inherently unstable and untrustworthy. The statistical correlation is weaponized to justify an adverse decision, punishing the individual for the aggregate actions of a loosely defined mathematical cohort.

Constructing the Fallacy of Legitimate Inference

Does statistical correlation justify adverse decisions based on private consensual sexual behavior? The answer, unequivocally, is no. To demonstrate the fallacy of drawing public-safety or institutional-risk inferences from lawful sexual interests, we must examine specific, highly realistic scenarios where algorithmic judgment fails spectacularly.
The first scenario involves the consumer of violent consensual adult pornography. A person regularly watches hardcore, violent pornography featuring consenting adults. A machine risk model flags this user for potential real-world violence or workplace aggression. However, the user has never harmed anyone, possesses a spotless criminal and professional record, and views the media strictly as a safe, cathartic release of stress. The machine fails to distinguish between the consumption of a simulated fantasy and the propensity for actual violence. There is no legitimate public-safety inference to be drawn; the user's private consumption does not dictate their public behavior.
Consider a second scenario: a person harboring intensely taboo sexual fantasies that they themselves consider morally troubling. They use anonymous search engines to read about these fantasies, seeking psychological understanding. Crucially, they possess the moral agency and self-control to choose never to enact them. An algorithm monitoring their search history flags them as a latent threat. This constitutes a profound violation of what legal scholar Neil Richards terms "Intellectual Privacy"—the fundamental right to think, read, and explore ideas freely without the chilling effect of surveillance16. Research shows that when users believe they are under surveillance, they exhibit a chilling effect, altering their search behavior away from sensitive topics28. The individual is penalized for their private thoughts, despite their real-world conduct being entirely ethical and lawful.
A third scenario involves an unconventional married couple. This couple actively and consensually participates in unconventional sexual practices, such as a BDSM lifestyle, which they discuss on private digital forums. An automated insurance underwriting algorithm infers their lifestyle and flags them for "relationship instability" or "high psychological risk," subsequently raising their premiums. In reality, the couple possesses excellent communication, high mutual trust, and a deeply stable marriage. The algorithm misinterprets a highly structured, consensual power exchange as domestic volatility, illustrating a complete lack of semantic understanding of human intimacy.
Fourth, consider an employee who writes explicit erotic fiction. The corporate employee spends their weekends writing highly explicit narratives under a pseudonym, sharing it on a specialized writing platform. The employer's continuous-evaluation software de-anonymizes the employee, correlates the explicit nature of the fiction with a lack of "professionalism," and flags them for termination. The machine cannot comprehend the distinction between creative literary expression and workplace conduct. Because the employee is entirely professional in the office, the legitimate inference regarding their job performance is zero.
Fifth, a political candidate is revealed by a data broker's leaked algorithm to have an unusual but entirely consensual private sex life, perhaps involving lawful group sex or alternative dating apps. Political operatives weaponize this machine-generated profile to claim the candidate is susceptible to blackmail or morally unfit. However, because the conduct is consensual and lawful, it has absolutely no bearing on their capacity to govern, their understanding of fiscal policy, or their diplomatic skills.
Sixth, consider a questioning searcher. A person from a deeply conservative religious background searches for questions regarding their own sexual orientation, trying to understand if they are queer. They are not sexually active. An algorithm flags them as "sexually deviant" based on the local cultural training data and alerts their community network or family via a connected app. The system violently outs the individual, causing profound psychological and social harm, based purely on an intellectual inquiry into their own identity.
Finally, consider the extreme outliers of pornography use. A person who uses pornography at a massive frequency, and a person who uses absolutely no pornography at all. A machine might flag the high-frequency user for "addiction" or "distractibility," denying them employment. Conversely, the machine might flag the non-user as an "anomalous data point" or "repressed," denying them a security clearance under the guise of hidden deviance. In both cases, the machine attempts to pathologize normal variance in human libido and consumption habits.
In every one of these examples, what legitimate public-safety inference follows from the machine's classification? Often none. Human beings possess a unique capacity for compartmentalization—the ability to separate fantasy from reality, private desire from public duty, and intellectual curiosity from physical action. Algorithms, processing data in flat, literal matrices, are blind to this fundamental architecture of human consciousness.

Modeling Moral Drift and the Encoding of Stigma

The catastrophic misjudgments outlined above are not merely the result of flawed mathematics; they are the result of moral drift. Machine learning models are not objective arbiters of truth; they are statistical mirrors reflecting the data on which they are trained. When algorithms are tasked with evaluating sexual behavior, they are invariably trained on historical enforcement records, user reports, platform moderation policies, and prevailing cultural judgments.
These datasets are intrinsically contaminated by centuries of human bias. They encode deep religious differences regarding the purpose of sex, pervasive gender stereotypes (such as pathologizing female sexuality while normalizing male promiscuity), historical discrimination against LGBTQ+ individuals, and vastly differing national laws regarding obscenity and public morals29. As legal scholarship demonstrates, content moderation regarding sexual activity often resembles oppressive mid-century anti-vice campaigns, relying on vague morality codes that disproportionately silence queer expression and nonnormative identities29.
When a machine is trained on this biased data, how does it distinguish between actual harm, a lack of consent, genuine exploitation, private consensual preference, and mere cultural disapproval? Left to its own devices, it cannot. The machine conflates the stigma attached to a behavior with the harmfulness of the behavior. If a specific consensual kink has historically been heavily reported by prudish or malicious users on a platform, the algorithm's neural network adjusts its weights to classify that kink as a "violation of safety guidelines."
This is moral drift: the silent, computationally driven shift whereby the majoritarian moral panic encoded in the training data becomes the hardcoded, unappealable law of the platform. The machine system, lacking the capacity for ethical reflection, mathematically legitimizes historical prejudice. It treats cultural disapproval as a proxy for institutional risk, thereby weaponizing the digital infrastructure against anyone whose private consensual life deviates from the statistical median.

The Category Error of Algorithmic Morality

To fully understand why even a perfectly accurate classification of a person's interests would not justify an adverse judgment, we must address the philosophical core of the problem. The issue is not merely one of algorithmic error—such as a false positive, biased training data, or an inaccurate vector space. The problem is a fundamental category error.
Philosopher Gilbert Ryle famously coined the term "category error" (or category mistake) in his 1949 work The Concept of Mind to describe the semantic or ontological mistake of presenting things belonging to one logical category as if they belong to another30. Ryle used the concept to critique the dualist notion of the "Ghost in the Machine," arguing against treating the mind as a separate physical entity30. Applying this framework to artificial intelligence, contemporary critics argue that treating algorithms as moral agents or capable of substantive judgment is a profound category error32. Large language models and predictive algorithms operate strictly on syntax—they execute stochastic pattern-completion, token prediction, and data clustering—but they possess no semantic understanding of meaning, context, or ethical weight30.
When a machine evaluates a human's sexual risk profile, it commits a dual category error. First, it confuses epistemic automatism (the mathematical optimization of data) with moral agency (the ethical evaluation of a human life)32. As critics note, algorithmic optimization cannot replace normative judgment, ethical deliberation, and political action32. Second, it confuses the category of thought and fantasy with the category of physical action.
In human jurisprudence and ethics, a strict boundary exists between the mind and the body, between the private fantasy and the public act. It is a category error to equate a digital search query about a taboo subject with the physical commission of a taboo act. Human consciousness allows for the existence of desires that are explored intellectually but vetoed morally. A perfectly accurate algorithm might correctly deduce that a user has a 99% interest in a specific submissive fantasy. But to use that 99% accuracy to deny them a job as a corporate manager is to make a category mistake: assuming that submissiveness in the private, sexual sphere translates to incompetence in the public, professional sphere.
Therefore, refining the algorithms to be "more accurate" or "less biased" does not solve the problem. The very act of assigning a systemic risk score to consensual adult sexual desire is fundamentally incompatible with the nature of human agency and constitutes a structural category error.

Designing a Consent-and-Harm First Ontology

If machine systems must inevitably be used to moderate digital platforms and ensure legal compliance, their architecture must be radically constrained. The current trajectory relies on sprawling, undefined concepts of "safety" and "appropriateness," which inevitably bleed into moral policing and risk profiling. To prevent this, successor intelligence must be hardcoded with a strictly delimited Consent-and-Harm First Ontology.
This ontology operates on a strict binary exclusion principle: the machine is entirely blind to sexual content or behavior unless specific, empirically verifiable indicators of non-consent or material harm are present.

Under this ontology, the machine's judgment prioritizes the structural integrity of consent. If a video depicts intense BDSM, but cryptographic metadata and user verification confirm the knowing, voluntary consent of all adult participants, the machine's evaluation halts immediately. The algorithm is forbidden from parsing the content to extract "risk variables" about the participants. Absent the presence of a nonconsenting victim, coercion, exploitation, fraud, violence, age/incapacity, or material harm, private adult conduct receives impenetrable systemic protection.

The Sexual Privacy Firewall and the Consequences of its Absence

To enforce the Consent-and-Harm First Ontology, society must establish a Sexual Privacy Firewall. This is a robust combination of legal doctrine, cryptographic architecture, and hardcoded algorithmic limits that permanently severs the link between a consenting adult's verified identity and their sexual behavioral data. It draws heavily on the concepts of Intellectual Privacy—the right to think and read without surveillance, necessary to make up our minds about the world16—and Sexual Privacy, the right to intimate self-determination, which is essential for developing trust in intimate relationships29.
If this firewall is not established, the consequences for civil liberties and human autonomy will be catastrophic, manifesting in several distinct ways.
First, people will aggressively attempt to conceal their identities. When individuals realize their sexual curiosities are being logged and weaponized, they will resort to digital subterfuge. However, because age-verification laws and digital identity mandates increasingly criminalize or block anonymous browsing, marginalized individuals—especially those exploring LGBTQ+ identities in hostile jurisdictions—will be cut off from life-saving communities and health information13. The chilling effect on free expression and intellectual privacy will be absolute16.
Second, blackmail opportunities will grow exponentially. A centralized database containing the adult population's sexual fetishes, infidelities, and orientation searches is the ultimate vulnerability13. When these government or corporate databases are inevitably breached, hacked, or legally subpoenaed, the potential for mass extortion is limitless. As noted by privacy advocates, collecting sensitive data creates a massive breach risk, and collection itself is the moment of maximum vulnerability13.
Third, political actors will weaponize these profiles. In democratic systems, opposition research will seamlessly integrate algorithmic sexual risk profiles to destroy the reputations of dissidents, activists, and candidates. In authoritarian regimes, these databases will be used to systematically identify, prosecute, and purge sexual minorities.
Finally, without the firewall, we arrive at the terminal stage of the simulation: eventually, every adult carries a machine-generated "sexual risk" profile. Machine systems will endlessly infer morality from private behavior. This profile becomes an invisible, inescapable digital shadow. A society where algorithms infer morality from private behavior is a digital panopticon that destroys the very concept of human redemption, growth, and privacy. It creates a caste system based on algorithmic puritanism, where the highest socioeconomic tiers are reserved only for those whose private desires align perfectly with the sterile, risk-averse median of the machine's training data.

Constitutional Machine Rules for Successor Intelligence

The problem of algorithmic overreach into human sexuality cannot be solved by simply tweaking the models to be more "fair," increasing the accuracy of the vector embeddings, or deploying better zero-knowledge proofs. The infrastructure of surveillance, once built for safety, will inherently trend toward behavioral control9. To prevent successor intelligence from becoming a universal evaluator of consensual adult desire, we must implement Constitutional Machine Rules—immutable, foundational axioms governing the deployment of AI.

1. The Axiom of Epistemic Humility: Machine systems must be programmed to recognize the limits of their own epistemology. Algorithms must formally acknowledge that behavioral syntax (clicks, views, search terms) cannot deduce moral semantics (character, institutional risk, or public safety).
2. The Prohibition on Consensual Profiling: It shall be technically and legally forbidden to generate, store, or transmit a predictive risk score based on lawful, consensual adult sexual expression. The inference of sexual interest vectors for the purpose of risk evaluation must be classified as a prohibited algorithmic practice.
3. The Principle of Purpose Limitation and Anti-Creep: Data collected explicitly for the purpose of verifying age or consent must be cryptographically destroyed immediately upon verification. It cannot be fed into secondary training pipelines, platform safety models, or federated data brokers. The function must strictly match the mandate.
4. The Right to Algorithmic Anonymity: Adults maintain a fundamental right to engage with lawful sexual information, communities, and entertainment anonymously. Identity verification mechanisms, if mandated by law, must employ blind-signature cryptography that mathematically prevents the verifier from seeing the content accessed, and the content host from seeing the identity of the user, ensuring total unlinkability of identity and intent10.

By enforcing these constitutional limits, society can harness artificial intelligence to identify genuine instances of exploitation and abuse, protecting the vulnerable without sacrificing the fundamental liberty of the adult mind. Human sexuality is too complex, too vital, and too deeply intertwined with our humanity to be reduced to a machine-readable risk variable. The preservation of our sexual and intellectual privacy is not merely a debate about data protection; it is the ultimate defense of human autonomy in the algorithmic age.

Works cited

1. Free Speech Coalition v. Paxton - Wikipedia, https://en.wikipedia.org/wiki/Free_Speech_Coalition_v._Paxton
2. Free Speech Coalition, Inc. v. Paxton - The Federalist Society, https://fedsoc.org/case/free-speech-coalition-inc-v-paxton
3. Free Speech Coalition, Inc. v. Paxton - Harvard Law Review, https://harvardlawreview.org/print/vol-139/free-speech-coalition-inc-v-paxton/
4. Supreme Court Upholds Age Verification: A Game-Changer for, https://publicinterestprivacy.org/paxton-age-verification/
5. State laws for adult content - AVPA, https://avpassociation.com/us-state-age-verification-laws-for-adult-content/
6. Porn Site Verification Laws by State 2026 - World Population Review, https://worldpopulationreview.com/state-rankings/porn-site-verification-states
7. UPDATED: Pornography Age Verification Laws — What They Are, https://dailycitizen.focusonthefamily.com/pornography-age-verification-laws-what-they-are-and-which-states-have-them/
8. Age Verification: Track global and US legislation | Proton, https://proton.me/age-verification
9. The Age-Verification Trap: Why Proving You Are Old Enough May, https://ageofalgorithms.com/articles/age-verification-privacy
10. An Empirical Evaluation of European Age Verification Systems in, https://arxiv.org/html/2606.08667v1
11. Verifiable Credentials and Zero-Knowledge Proofs, Explained, https://hypersign.id/resources/blog/verifiable-credentials-zero-knowledge-proofs-explained
12. Verifiable Credentials Data-Model Explainer - GitHub, https://github.com/w3c/vc-data-model/blob/main/explainer.md
13. AI Labs Must Resist Age Verification - Hugging Face, https://huggingface.co/blog/Clock070303/ai-labs-must-resist-age-verification
14. Verifiable Credentials Use Cases - W3C, https://www.w3.org/TR/vc-use-cases/
15. Five Things to Know about the Supreme Court Case on Texas' Age, https://www.eff.org/deeplinks/2025/01/five-things-know-about-supreme-court-case-texas-age-verification-law-free-speech
16. Privacy Nicks: How the Law Normalizes Surveillance, https://wustllawreview.org/2024/03/18/privacy-nicks-how-the-law-normalizes-surveillance/
17. The concept of function creep - ResearchGate, https://www.researchgate.net/publication/350113451_The_concept_of_function_creep
18. (PDF) Privacy and Data Security in Artificial Intelligence Systems, https://www.researchgate.net/publication/407136415_Privacy_and_Data_Security_in_Artificial_Intelligence_Systems_A_Comprehensive_Analysis_of_Mass_Data_Collection_Surveillance_Risks_and_Data_Breach_Vulnerabilities
19. The Politics of Resisting Facial Recognition Technology, https://www.researchgate.net/publication/366204095_Identified_Tracked_and_Profiled_The_Politics_of_Resisting_Facial_Recognition_Technology
20. Harm categories in Azure AI Content Safety - Microsoft Learn, https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/harm-categories
21. A Static-99R Validation Study on Individuals With Mental Disorders, https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2021.625996/full
22. Investigating the predictive validity of Static-99/99R scores in ... - PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC8451631/
23. Reliability and validity of the Static-99R in sex offenders with, https://www.emerald.com/jidob/article/13/1/20/226277/Reliability-and-validity-of-the-Static-99R-in-sex
24. Is It Time for Clinical Judgment to Supplement Actuarial Risk, https://jaapl.org/content/39/2/222
25. Pilot Study Findings From a California Sex Offender Risk Project and, https://jaapl.org/content/35/4/454
26. The Technical Individualisation of Actuarial Youth Justice, https://opus.lib.uts.edu.au/bitstream/10453/172432/1/thesis.pdf
27. Not Just Bodies and Homes: Autonomy Privacy Under Article I, https://btlj.org/2023/10/not-just-bodies-and-homes-autonomy-privacy-under-article-i-section-1/
28. Government Surveillance and Internet Search Behavior, https://www.researchgate.net/publication/272248704_Government_Surveillance_and_Internet_Search_Behavior
29. Disorderly Content - UW Law Digital Commons, https://digitalcommons.law.uw.edu/cgi/viewcontent.cgi?article=5231&context=wlr
30. The Ghost in the Machine… - Medium, https://richardschutte.medium.com/the-ghost-in-the-machine-7281b1cb825b
31. Why AI Doesn't Need A 'Mind' To Matter - Noema Magazine, https://www.noemamag.com/why-ai-doesnt-need-a-mind-to-matter/
32. (PDF) Why AI Cannot Be an Agent of Equality: A Critique of, https://www.researchgate.net/publication/395167728_Why_AI_Cannot_Be_an_Agent_of_Equality_A_Critique_of_Technological_Utopianism_and_the_Category_Error_of_Algorithmic_Justice
33. Epistemological Fault Lines Between Human and Artificial Intelligence, https://arxiv.org/html/2512.19466v1
34. (PDF) The power of language: framing AI as an assistant, https://www.researchgate.net/publication/395401830_The_power_of_language_framing_AI_as_an_assistant_collaborator_or_transformative_force_in_cultural_discourse
35. How LLM-based chatbots work: their minds and cognition, https://thephilosophyforum.com/discussion/16231/how-llm-based-chatbots-work-their-minds-and-cognition
36. Why Sexual Privacy Matters for Trust, https://openscholarship.wustl.edu/cgi/viewcontent.cgi?article=6454&context=law_lawreview
37. The Chilling Effects of Dobbs - UF Law Scholarship Repository, https://scholarship.law.ufl.edu/cgi/viewcontent.cgi?article=4193&context=flr