{"id":177079,"date":"2026-03-21T21:11:53","date_gmt":"2026-03-22T05:11:53","guid":{"rendered":"https:\/\/lukeford.net\/blog\/?p=177079"},"modified":"2026-03-22T16:09:54","modified_gmt":"2026-03-23T00:09:54","slug":"when-i-hear-people-with-power-abuse-safety-i-reach-for-my-keyboard","status":"publish","type":"post","link":"https:\/\/lukeford.net\/blog\/?p=177079","title":{"rendered":"When I Hear People with Power Abuse “Safety,” I Reach for My Keyboard"},"content":{"rendered":"

There is a tell. When powerful institutions want to expand their authority, restrict your choices, suppress a competing viewpoint, or insulate a decision from challenge, they reach for one word before any other. The word is “safety”. It ends arguments. It converts institutional preference into moral necessity. It makes the person asking questions look like the problem.
\nThis essay is about what that word is actually doing in 2026 America, and who pays when it does it.
\nStephen Turner’s<\/a> work on expertise shows that the most durable forms of cognitive authority are those that present coalition interests as technical necessity. Safety is the word that performs this conversion most efficiently. It borrows the legitimacy of the physician reporting a test result, the engineer calculating a load, and deploys it in service of decisions that are fundamentally political: who gets to define the risk, whose harms count, and who absorbs the cost when the definition is wrong. The people who absorb that cost are almost never the people who set the definition.
\nWhat follows is ten cases. Each one follows the same structure. An institution invokes safety. The definition of safety it uses serves its own interests. The risks it elevates justify its authority and protect it from liability. The risks it suppresses are the ones its interventions create. The costs of that suppression fall on people without the resources to contest the definition. And the institution is insulated from accountability because safety, by construction, cannot be questioned without appearing to endanger people.<\/p>\n

Emergency Psychiatric Holds<\/a><\/strong><\/p>\n

The duty to protect is one of the most powerful phrases in American clinical practice. A clinician who identifies an ambiguous risk of self-harm can initiate an involuntary hold that removes a person’s freedom, places them in an institutional setting, and subjects them to evaluation and treatment without consent. The threshold for initiating this process is low. The threshold for contesting it, for a person in psychiatric distress without money, a lawyer, or institutional backing, is nearly impossible to meet.
\nThe safety claim here is real in its intention and often catastrophically wrong in its effect. Adolescent psychiatric inpatient settings frequently retraumatize through restraint, isolation, and forced procedures, particularly in young people with prior trauma histories. The institutional dynamics mimic the dynamics of past abuse: loss of control, coercive physical contact, isolation from support, and the redefinition of any protest as a symptom. Post-discharge suicidality spikes have been documented in multiple review contexts. None of this is tracked systematically, because the system is not designed to track the harms it produces. It is designed to track the harms it prevents, which it measures against a counterfactual it controls.
\nWhen the person subjected to this describes what happened as harmful, the institution has a ready response: you lack insight into your own condition, your protest is part of the disorder, and we acted to protect you. The recursive loop is closed. The complaint becomes evidence that the intervention was necessary.<\/p>\n

School-Based Threat Assessments<\/strong><\/p>\n

Zero-tolerance discipline and behavioral threat assessment programs operate on a safety logic that prioritizes the prevention of low-probability catastrophic events over the routine welfare of the students subjected to them. A student who makes an ambiguous comment, draws something disturbing, or behaves in ways that trigger an administrator’s pattern-matching gets flagged, searched, referred to law enforcement, or suspended. The school has protected itself from the worst-case scenario. The student absorbs a permanent disciplinary record, academic disruption, and in some cases a pathway into the criminal legal system.
\nThe populations who bear this cost are not random. The safety the institution is protecting is partly its own: liability exposure, reputational risk, and the administrative defensibility of having responded to a potential threat. The safety of the student, measured in educational outcomes and long-term development, is not the primary variable in the threat assessment matrix.
\nWhen parents contest these decisions, they encounter a system designed to be contested only from outside. The people inside it speak the language of professional risk assessment. The parent who says my child is being harmed by this process is told that the process exists to protect children, which is a different claim, and one that does not engage the specific harm being described.<\/p>\n

“Trauma-Informed” Expansion of Coercive Tools<\/a><\/strong><\/p>\n

Trauma-informed care began as a genuine clinical insight: institutions that work with traumatized people should understand trauma, recognize its manifestations, and avoid replicating traumatizing conditions. That insight was then captured by the institutional machinery it was meant to reform. The vocabulary of trauma became a justification for expanding the very coercive tools it was meant to constrain.
\nIn youth residential facilities, the trauma-informed label now covers restraints, isolation rooms, forced compliance procedures, and intensive behavioral monitoring. The framing is that these measures are delivered with sensitivity and awareness of trauma history, which makes them different from the crude coercion they resemble. For the young person experiencing them, the difference is not always apparent. For the institution, the label provides cover: adverse outcomes are framed as the necessary cost of treatment, as evidence of the severity of the underlying condition, or as the result of inadequate family support. They are not framed as evidence that the protocol is wrong.
\nThe research base for many of these interventions is thin. Adverse event reporting is structurally weak. Long-term outcome comparisons with less coercive alternatives are largely absent, because funding flows toward the institutional model and away from the alternatives. This is Turner’s information deprivation mechanism operating at the design stage of the research enterprise: the data that would challenge the coalition’s safety claims is not collected, because the coalition controls what gets funded.<\/p>\n

Content Moderation and “Misinformation Safety”<\/strong><\/p>\n

During the Covid period, platform moderation and its government partners developed a working definition of misinformation: information that deviated from guidance issued by public health agencies. The definition was circular. The agencies defined what was true. Deviation from what the agencies said was therefore false. Content that was false was therefore harmful. Content that was harmful could be suppressed in the name of safety. The people who set the definition and the people who enforced it were participants in the same coalition.
\nThe specific content suppressed under this regime included, at various points, discussion of natural immunity, questions about mask efficacy at the community level, early reporting on the possibility of a lab origin for the virus, and clinical observations from physicians that contradicted agency guidance. Some of this content was wrong. Some of it was right, and turned out to be more accurate than the guidance it contradicted. The suppression did not distinguish between these cases, because the definition of misinformation was not designed to track truth. It was designed to track deviation from the coalition’s position.
\nThe people most harmed by this regime were not the well-resourced commentators who could find alternative platforms and audiences. They were the individual clinicians, the community health workers, and the ordinary people trying to evaluate competing claims about their own health decisions, who found that the information environment had been curated in ways they could not see, by a coalition they could not identify, using criteria they were not permitted to contest.
\nThe safety claim here is about protecting people from false information that leads to harmful decisions. This is a real concern. But a regime that cannot be contested from outside, that defines its own errors out of the category of error, and that suppresses accurate information along with inaccurate information in the name of safety, is not protecting anyone from harm. It is protecting the coalition from accountability.<\/p>\n

AI Safety Standards That Entrench Incumbents<\/strong><\/p>\n

The alignment-focused AI safety coalition has achieved something remarkable: it has framed the question of who is allowed to build advanced AI systems as a safety question. Compute caps, pre-deployment audits, licensing requirements, and mandatory safety evaluations are presented as protections against existential risk. They are also barriers to entry that only large, well-capitalized organizations can clear.
\nA small research team or independent developer who wants to train a competitive model faces compliance costs that are not proportional to any plausible safety risk from their specific work, but are proportional to the regulatory burden that the largest incumbents helped design and can absorb. The safety standard, once institutionalized, functions as a market structure: it determines who is permitted to compete and who is not. The people who designed the standard are the people it most benefits.
\nThe risks from AI development that are elevated in this framework are the speculative catastrophic ones: misalignment, existential threat, loss of human control over powerful systems. The risks that are suppressed are the concrete immediate ones: concentration of AI capability in a small number of organizations, labor displacement without social support, deployment of biased systems in consequential decisions, and the epistemic effects of AI systems trained on human content reshaping the information environment. These suppressed risks are not less real than the elevated ones. They are less useful to the coalition that controls the safety definition, because addressing them would require constraining the coalition’s own members rather than constraining their competitors.<\/p>\n

Public Health Mandates with Narrow Risk Framing<\/strong><\/p>\n

The logic of public health safety during the Covid period elevated one category of risk, infectious spread, to a level that justified interventions across the full range of social life. School closures, business restrictions, isolation protocols, and behavioral mandates were all safety measures by the coalition’s definition. The harms produced by these measures, delayed cancer diagnoses, educational loss, mental health deterioration, developmental disruption for young children, domestic violence, and economic devastation for people without savings or remote-work capability, were categorized as the costs of safety rather than as safety failures.
\nThis is Turner’s risk selection mechanism in its clearest form. The coalition elevated the risks it could address and suppressed the risks its interventions created. The costs of that suppression fell disproportionately on people with the fewest resources to adapt: parents without childcare alternatives, workers without sick leave, students in households without reliable internet, small business owners without capital reserves. The people designing the interventions could work from home, send their children to private schools that remained open, and access healthcare through channels that remained functional. The definition of safety they applied was not unreasonable on its own terms. It was unreasonable in its narrowness, and the narrowness was not random.<\/p>\n

Child Protection and Family Court<\/strong><\/p>\n

The phrase safety of the child operates in the family court system the way safety operates everywhere else in this map: it preemptively justifies intervention, places the burden of proof on the person contesting it, and insulates the intervening institution from accountability. The threshold for initiating a child protective investigation is low, as it should be in cases of genuine danger. The threshold for resolving it in a parent’s favor, especially for parents without money, legal representation, or social capital, is much higher.
\nIn practice, families targeted by child protective interventions are not random. The removal of a child from a family is among the most severe interventions the state can make in a person’s life. It is authorized by a safety standard set by professionals who are not accountable to the family, evaluated by a court system that frequently treats the agency’s assessment as authoritative, and contested through a legal process that the agency has navigated thousands of times and the family typically encounters for the first time.
\nThe 2026 Supreme Court rulings on parental rights and state-level restrictions on the use of child protection proceedings in family disputes over medical and educational decisions represent exactly the kind of coalition shift Turner describes. The prior coalition’s admissible reality, that professional assessment of child safety should override parental authority in the relevant cases, is now being contested by a different coalition with different institutional backing. The underlying question, what actually protects children, has not been answered. The question of who gets to define the answer has shifted.<\/p>\n

Workplace “Psychological Safety” Systems<\/strong><\/p>\n

Psychological safety as a concept in organizational research described the conditions under which people feel comfortable taking interpersonal risks, speaking up, and raising concerns without fear of punishment. It was a genuinely useful insight about how high-functioning teams operate. It was then captured by HR compliance machinery and deployed as a speech governance tool.
\nIn its institutional form, psychological safety means something specific: the organization has created a reporting structure through which employees can flag conduct that makes them feel unsafe. The conduct flagged is not limited to discrimination or harassment in any legally defined sense. It encompasses disagreement expressed in ways that a colleague experiences as threatening, viewpoints that make someone uncomfortable, and speech that challenges the coalition’s preferred moral framework on topics like gender, race, or political economy.
\nThe enforcement of psychological safety in this form is asymmetric. Lower-status employees, those without tenure, institutional backing, or social capital, are more exposed to complaints than higher-status ones. The person whose speech is flagged has limited ability to contest the characterization. The system is designed to take complaints seriously, which is a reasonable principle in cases of genuine misconduct and a mechanism for selective enforcement in cases where the definition of misconduct is controlled by one faction within the organization. The word safety insulates the enforcement mechanism from challenge: questioning the process looks like opposing safety.<\/p>\n

Campus and Community “Safe Space” Expansions<\/strong><\/p>\n

The safe space concept began as a specific claim about specific populations, that people with particular histories of marginalization benefit from environments where they can discuss their experiences without encountering hostility. This is defensible and sometimes important. It was then generalized into a principle that discomfort constitutes harm, that exposure to disagreeable viewpoints is a safety issue, and that institutions have an obligation to restrict speech that causes discomfort to members of designated groups.
\nThe consistent effect of this generalization has been to constrain the range of arguments that can be made in institutional settings, to provide a mechanism for removing speakers, restricting events, and sanctioning faculty and students whose views fall outside the coalition’s admissible range, and to make that constraint unchallengeable because challenging it is itself framed as a safety threat. The people most affected by this constraint are not those within the coalition who benefit from the expanded definition of harm. They are those who want to make arguments the coalition has pre-classified as harmful, regardless of the quality of the arguments.
\nJonathan Haidt and Greg Lukianoff’s work on safetyism, the treatment of discomfort as danger, is relevant here not as a political critique but as a developmental one. The evidence that exposure to manageable adversity, including disagreement, criticism, and challenge, builds resilience is substantial. The evidence that institutional protection from discomfort builds resilience is thin. If the safety claim is taken seriously on its own terms, the question of whether these regimes make people safer over time, rather than more fragile and more dependent on institutional protection, is exactly the question they are structured to not ask.<\/p>\n

Financial “Safety and Soundness” De-risking<\/strong><\/p>\n

Banks and payment processors have expanded their use of safety and soundness standards to justify restricting accounts and transactions associated with politically disfavored industries, organizations, or individuals. The mechanism is usually called de-risking: the institution determines that certain customers or transaction types pose reputational or regulatory risk and terminates or refuses the relationship. The decision is framed as a prudential safety judgment. It is rarely subject to external review, because financial institutions have broad discretion in customer selection and the safety framing insulates that discretion from challenge.
\nThe industries and organizations most affected by this process are those that the regulatory coalition has pre-classified as risky: certain firearms-related businesses, cannabis operations in states where it is legal, politically disfavored advocacy organizations, and individuals who have been publicly associated with contested positions. These are not random. They are the categories the coalition controlling the safety definition has decided to suppress through the mechanism of financial access. Small organizations without the resources to maintain multiple banking relationships or navigate the legal system are most exposed. Large organizations can absorb the cost or find alternatives.
\nThe safety claim here is not entirely empty. There are genuine risks in some of the de-risked categories. But the definition of risk being applied is not a technical assessment of financial exposure. It is a political judgment about which activities are acceptable, dressed in the language of prudential regulation. The person or organization that loses banking access has no meaningful recourse, no right to a hearing, and no mechanism to contest the characterization of their activity as risky. They have been placed outside the system by a safety determination they cannot see, cannot challenge, and cannot appeal.<\/p>\n

The Compression<\/strong><\/p>\n

Across all ten cases, the mechanism is the same. The institution elevates certain risks, the ones its interventions address and that justify its authority. It suppresses competing risks, the ones its interventions create and that would undermine its authority. It uses the word safety to convert those choices into moral imperatives, so that questioning the definition is reclassified as endangering people. And it enforces the definition through systems that are accessible to people with resources and largely impenetrable to people without them.
\nThe costs fall consistently on the same populations: people without money, legal representation, institutional backing, or social capital. The adolescent in the psychiatric hold without an advocate. The student in the threat assessment without a parent who knows how to navigate the system. The family in the child protective proceeding without a lawyer. The small business owner without a compliance department. The individual whose account is closed without explanation or appeal.
\nTurner’s framework names what these cases have in common. The coalition that defines safety also controls the evidence about whether its definitions work, designs the metrics that evaluate its own interventions, and manages the information environment in which challenges to its authority appear. When the definition produces harm, that harm is not admissible as evidence against the definition. It is absorbed, reframed, and used to justify more of the same.
\nThe investigation priority is always the same. Demand standardized adverse event data. Require preregistered long-term outcome tracking. Insist on comparison to alternatives. Ask whose risks are being counted and whose are not. Ask who bears the cost when the definition is wrong, and whether anyone bears the cost of having set it.
\nWhen people with power invoke safety in ways that expand their authority, restrict your choices, and insulate their decisions from challenge, the question is not whether safety matters. Of course it does. The question is whether the invocation is tracking actual harm to actual people, or whether it is tracking the coalition’s interest in remaining the coalition. In 2026, across the cases in this map, the answer is far too often the latter.
\nWhen I hear people with power invoke safety in that way, I reach for my keyboard.<\/p>\n","protected":false},"excerpt":{"rendered":"

There is a tell. When powerful institutions want to expand their authority, restrict your choices, suppress a competing viewpoint, or insulate a decision from challenge, they reach for one word before any other. The word is “safety”. It ends arguments. … Continue reading →<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[43163],"tags":[],"class_list":["post-177079","post","type-post","status-publish","format-standard","hentry","category-safety"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/posts\/177079","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=177079"}],"version-history":[{"count":7,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/posts\/177079\/revisions"}],"predecessor-version":[{"id":177393,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=\/wp\/v2\/posts\/177079\/revisions\/177393"}],"wp:attachment":[{"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=177079"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=177079"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/lukeford.net\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=177079"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}