“Sex offender” is a legal classification, not a diagnosis or a coherent human type. Michael Caruso’s reported reasoning exposes how stereotypes can obscure both individualized risk after conviction and the trusted relationships through which sexual harm often occurs before anyone is ever on a registry.
In February 2026, months before Palm Beach County Clerk of the Circuit Court and Comptroller Michael Caruso was arrested, law enforcement arranged for a conversation with him to be recorded.
According to the arrest affidavit incorporated into Gov. Ron DeSantis’s suspension order, Caruso argued that the accusations against him were “impossible.” His reasoning, as investigators summarized it, was that sexual deviation is directed toward one sex and that because he was attracted to women, he therefore could not be a sex offender. The affidavit also records another statement attributed to him: that if the matter were reported to authorities, there would be little of him left.
That is more revealing than an ordinary denial.
This is a condensed representation of the logic investigators attributed to Caruso, not a verbatim quotation.
Caruso was arrested on August 18 on five felony counts: kidnapping, lewd or lascivious molestation, lewd or lascivious exhibition, luring or enticing a child, and child abuse causing mental injury. Gov. Ron DeSantis immediately suspended him from office. The charges remain allegations. There has been no adjudication of guilt, and Caruso is entitled to the criminal process.
This article is not an attempt to decide that case.
It is about the idea contained in that recorded statement.
What, exactly, do Americans think a “sex offender” is? And how much public policy has been built around the picture that appears in our heads when we hear the noun?
In ordinary speech, “sex offender” often sounds as though it names a recognizable human type: a person with a particular psychology, a particular set of desires, a particular danger profile and perhaps even a particular way of moving through the world.
Legally, it does nothing so precise.
In Florida, for example, the term “sexual offender” is defined through statutory criteria tied largely to convictions for qualifying offenses and related legal conditions. Florida then attaches registration duties and public disclosure to that classification.
That is a legal category. It is not a clinical diagnosis.
And the conduct that can sit beneath sex-offense labels is remarkably heterogeneous. The Justice Department’s SMART Office distinguishes among people who abuse children, people convicted of rape, people involved in internet-facilitated offending and other groups, while warning that even conventional typologies have serious limitations and crossover between categories. It notes, among other things, that not everyone who sexually abuses a child has pedophilia and that online-only and contact offending do not present identical patterns of risk.
Can be serious and harmful while involving no direct physical contact with a victim. The mechanisms, access patterns and prospective risk questions may differ from hands-on abuse.
Involves direct physical access and raises its own questions about coercion, opportunity, relationship, authority, repetition and safeguarding.
Different does not mean harmless versus harmful. It means the same broad legal label can conceal materially different mechanisms and risk questions. Crossover can occur, which is another reason individualized assessment matters.
The relevant distinctions are not cosmetic. Conduct may be violent or nonviolent, coercive or non-coercive, contact or entirely non-contact. It may involve adults or children. A victim may be an acquaintance, intimate partner, family member or, very rarely in cases involving children, a stranger. Conduct may be isolated or repeated. It may arise from different motivations and different combinations of sexual interest, antisocial behavior, opportunity, access and other factors.
Different does not mean harmless.
“Different means different—and justice has no business pretending otherwise.”
Those distinctions matter because culpability is individualized. Treatment need is individualized. The mechanisms that lead to offending can differ. The circumstances that make future offending more or less likely can differ. Even the federal government’s review of sexual-offense risk assessment emphasizes that no single factor adequately predicts future offending and that contemporary assessment increasingly combines multiple static and dynamic factors rather than relying on a single category or intuition.
Yet one noun can make all that complexity disappear.
Before August 18, Michael Caruso was not the cultural image most Americans have been trained to associate with the phrase “sex offender.”
He had served in the Florida House beginning in 2018 and continued there until his appointment as Palm Beach County Clerk of the Circuit Court and Comptroller took effect in August 2025. He was a certified public accountant and former business owner. He was a county constitutional officer. He was a grandfather. He occupied positions carrying professional standing, political legitimacy and public trust.
None of that establishes innocence.
That is precisely the point.
The imagined sex offender is frequently an outsider. Someone socially marginal. Someone visibly suspect. Someone parents would know to keep away from their children.
Caruso represented almost the reverse of that picture.
If the affidavit accurately describes his thinking, he appears to have looked at his own identity—heterosexual man, public official, grandfather, established professional—and treated those characteristics as evidence that he could not belong to the category he associated with sexual offending.
But respectable biography is not a risk assessment.
Neither is marginality.
Neither is political affiliation, profession, family status, sexuality, wealth, dress, neighborhood, church membership or public reputation.
The question is what a person does.
That sounds obvious. Our language suggests that it is not.
The most important feature of the allegations may not be their location or even their severity. It may be the architecture of access.
According to the affidavit, the child was Caruso’s grandson. Investigators described incidents during a fishing trip and a family cruise. On the cruise, the affidavit says, Caruso separated the child from his parents under the stated pretext of taking him for ice cream and remained alone with him for at least an hour. The affidavit alleges sexual abuse during that period. It also describes other reported incidents and a later family confrontation in which Caruso made limited admissions about circumstances while denying the central sexual allegations.
Look at the structure, not merely the accusation.
There was no unknown man approaching a child at the edge of a playground.
The alleged pathway was already built.
Existing relationships meant proximity did not have to be engineered by a stranger; it came with the role.
Professional and political standing need not have caused anyone to lower their guard for those qualities to matter culturally. Respectability itself is one of the cues by which human beings decide whom to trust.
That is why “respectable people can also commit crimes” is too weak a lesson.
The more important lesson is that the characteristics that make someone appear least like the stereotypical offender can coexist with—and sometimes help sustain—the trust through which access becomes possible.
Research on sexual assault makes the broader point difficult to ignore. In the Bureau of Justice Statistics’ analysis of sexual assaults against juveniles reported to law enforcement, almost all identified offenders were known to the victim. Strangers accounted for only 7 percent overall, 5 percent for children ages 6 through 11, and just 3 percent for children younger than six. For the youngest children, stranger offending was exceedingly rare compared with abuse by people already within the child’s world.
In the BJS study, almost all identified offenders were already known to the juvenile victim.
This does not mean every trusted adult should be treated as suspicious.
It means trust cannot substitute for safeguarding.
Stranger-danger mythology can fail before conviction, too.
Registry politics often speaks a geographic language.
Where does this person live? How close is the house to a school? How close to a park? Is there a registrant in the neighborhood? Who is living within a particular radius of children?
That framework has intuitive power because it corresponds to a particular story about danger: an identifiable dangerous outsider obtaining physical proximity to potential victims.
The allegations against Caruso describe something almost opposite.
That distinction matters far beyond this case.
A public registry can tell a parent where certain previously convicted people report living. It cannot tell that parent which uncle, grandparent, family friend, coach, teacher, clergy member, professional or institutional authority who has never been convicted may pose a risk. It cannot map secrecy, grooming, family deference, authority, opportunity or an adult’s ability to create unsupervised access.
And the evidence on geographic controls is far less comforting than their intuitive appeal. The Justice Department’s review of residence-restriction research found multiple studies in which living near schools or day-care centers did not distinguish recidivists from nonrecidivists or in which restrictions did not produce measurable reductions in sexual recidivism. One study summarized by the SMART Office found that social proximity appeared more consequential than geographic proximity. The review also emphasizes limitations in the research rather than pretending the evidence is absolute.
The same federal review describes the broader evidence on registration and notification as mixed: some studies found benefits on certain outcomes, while others found no statistically significant effects, with methodological limitations complicating sweeping conclusions.
That is enough to support a narrower and more important proposition.
“A map of previously convicted people is not a map of where sexual harm originates.”
Confusing the two can distort prevention before anyone is ever arrested.
A society trained to identify a visible class of dangerous outsiders may become very good at looking outward while remaining comparatively inattentive to the ordinary mechanisms by which trusted people obtain privacy, authority and opportunity.
There is another reason Caruso’s reported statement is unusually revealing.
He was not institutionally naïve.
A Florida clerk of the circuit court does not adjudicate criminal guilt. Caruso was not a judge. But the office he headed sits extraordinarily close to the machinery by which accusations become cases and cases become durable public records. Florida court clerks maintain and manage access to court records, and Palm Beach County’s clerk system includes public access to criminal, civil and other court files.
Caruso had also spent roughly seven years in the Florida Legislature before becoming clerk. Gov. DeSantis’s appointment announcement itself recounted his legislative service from 2018 through 2025.
Worked inside the system that writes criminal and collateral rules.
Led an office responsible for records produced by criminal, civil and family-court processes.
Operated in a state where sex-offense registration creates extensive, highly visible consequences.
It would therefore be remarkable if a longtime Florida lawmaker and sitting county clerk were wholly unfamiliar with the fact that criminal labels acquire lives far beyond the moment a charge or sentence is imposed.
That is especially true in Florida.
The state maintains a publicly searchable sexual-offender and predator system. Florida law requires extensive registration information, and current FDLE materials describe lifetime registration requirements for sexual offenders and predators along with registration of home and work addresses and other identifying information.
We cannot know from his job titles what Caruso personally understood about every consequence of Florida registration law. We should not pretend we can.
But his positions placed him unusually close to systems in which names, charges, judgments, records and classifications matter enormously.
That leaves at least two plausible ways to understand the logic attributed to him.
Perhaps he sincerely thought in categorical terms: people like me are not sex offenders.
Or perhaps this was defensive reasoning from someone who understood, at least generally, how catastrophic acquiring that identity could become: whatever happened, I cannot be that.
We do not need to choose between those explanations.
Either reveals the enormous social force carried by the noun.
This is where the case becomes more than a story about one man’s strange reported defense.
Both substitute an imagined identity for individualized analysis.
The first substitutes an imagined identity for evidence about what happened.
The second can substitute a statutory identity for individualized evidence about what the person is like now, what produced the conduct, what treatment is needed and what future risk the person actually presents.
Neither follows automatically.
This does not mean past conduct becomes irrelevant the moment a sentence ends. Prior behavior can be highly relevant to future risk. Some histories are profoundly concerning. Repetition matters. Persistent sexual interests can matter. Prior offending matters. General criminality can matter. Age and other empirically supported factors may matter. The SMART Office’s risk-assessment review is emphatic on precisely this point: meaningful prediction requires combining relevant factors, not finding a single magical variable.
That is why individualized analysis is more serious, not less serious, than categorical thinking.
It asks harder questions.
Consider what disappears when legal and public language reduces a person to a single designation.
Repeated coercive hands-on abuse presents one set of questions.
A wholly non-contact offense presents another.
An offense arising from an established relationship of trust presents different prevention issues from an offense involving a stranger.
Persistent patterns differ from isolated conduct.
Image-based offending may involve different mechanisms and risk patterns from direct contact abuse, even though crossover exists and cannot simply be assumed away. The Justice Department’s own review recognizes these differences and reports that online-only groups and conventional contact groups do not present identical risk profiles.
Adolescent conduct, adult conduct, violent conduct, nonviolent conduct, coercive conduct and conduct without the same mechanism of coercion are not interchangeable simply because legislatures can place them beneath a common statutory umbrella.
This is not offense ranking by moral sentiment.
It is the basic proposition that different facts raise different questions.
The federal literature on risk assessment exists because researchers and practitioners already know that “sex offender” is not a sufficient prediction model. The SMART Office describes tools that combine multiple factors, notes that dynamic information can identify intervention targets, and explicitly recognizes the need to match treatment and management efforts to an individual’s risk and criminogenic needs rather than apply uniform interventions.
If a noun were enough, none of that work would be necessary.
The allegations against Caruso, as described in the affidavit, involve an especially grave constellation: a young family member, alleged hands-on sexual abuse, repeated incidents, existing access, isolation from other adults and what investigators describe as circumstances suggesting the use of trusted authority to obtain opportunity.
Another person can enter the broad legal category without sharing all—or even many—of those characteristics.
The fact that both may ultimately be called the same thing is not proof that the category conveys no information.
It is proof that the category does not convey enough.
Serious conduct makes this argument stronger, not weaker.
There is an understandable temptation to hear any discussion of heterogeneity as an attempt to minimize sexual harm.
That would be a mistake here.
The conduct described in the Caruso affidavit is grave. The alleged victim was a young child within his own family. The accusations describe repeated sexual acts and an extraordinary breach of trust. Nothing about arguing for individualized justice requires sanding those facts down.
The seriousness is exactly why the example matters.
If conduct carrying that collection of characteristics and a wholly non-contact, nonviolent offense can ultimately feed into the same broad public identity, then the identity itself plainly cannot communicate everything we need to know about culpability, mechanism, treatment or future danger.
A category can carry meaningful information without carrying sufficient information.
“Murder” tells us someone was convicted of causing a death under a particular criminal law. It still does not eliminate the need to know intent, circumstances, history and future risk.
Yet with sexual offenses, the grammatical move from act to identity has become unusually complete.
A person does not merely commit a sex offense.
He is a sex offender.
“The noun quietly swallows the person.”
Criminal law spends enormous energy distinguishing facts.
Those distinctions affect charging, proof, sentencing and treatment.
Then a remarkable thing can happen after conviction.
The justice system that insisted factual distinctions mattered enormously can place people into a post-conviction framework where the most publicly salient fact becomes a two-word identity:
SEX OFFENDER.
Florida’s registry makes the legal transformation literal. The state defines eligibility through qualifying convictions and then makes identifying information about people within that category available to the public through its registry.
What was once a detailed factual inquiry becomes a searchable noun.
That is an extraordinary compression of information.
And compression has consequences when policymakers or members of the public begin treating the compressed category as though it contains all the information that was discarded.
The paradox is difficult to escape:
The justice system can insist that factual distinctions matter enormously when determining guilt and punishment, then construct post-conviction policy in which many of those distinctions become secondary to the category itself.
None of this requires claiming that a person with a prior conviction can never pose a future risk.
That would merely replace one categorical fiction with another.
The problem is not that public information contains no useful information. It is that the public can be encouraged to mistake one particular dataset for a comprehensive model of danger.
A registry identifies people who have already passed through a qualifying legal process.
It does not identify the respected adult who has never been convicted.
It does not identify the family member whose access is built into everyday life.
It does not identify the coach who can create private moments, the clergy member protected by reverence, the professional protected by status, the family friend protected by familiarity or the institution whose reputation discourages uncomfortable questions.
The BJS findings are stark enough that they should change how we think about prevention, not merely how we describe crime after the fact. Among juvenile victims in the study, 93 percent of identified offenders were known to the child. For children under six, strangers accounted for only 3 percent. The dominant pattern was not an unknown outsider crossing into the child’s world. It was someone who was already there.
Caruso’s alleged circumstances make that tension unusually visible.
The state can tell a parent how far a registrant lives from a school.
But no map can tell that parent whether “Grandpa is taking him for ice cream” is safe.
That kind of safety requires something harder than geographic awareness.
It requires families and institutions to think about access.
About supervision.
About boundary violations.
About secrecy.
About adults who engineer isolation.
About whether status makes questioning someone feel socially forbidden.
About whether a child has realistic ways to disclose something involving a person everyone else trusts.
Public safety fails when we train people only to recognize monsters.
A more serious framework begins with four questions.
Culpability and proportional punishment. Serious, coercive, violent, repeated or trust-exploiting conduct can justify severe consequences.
Mechanism, treatment and prevention. Different combinations of sexual interest, antisociality, opportunity, cognition, access and other factors call for different responses.
Past conduct belongs in the inquiry, but prospective risk requires combinations of empirically relevant factors rather than a single label.
Management, supervision and treatment should be matched to actual risk and need rather than applied uniformly.
Those are four different questions.
“Sex offender” cannot answer any of them adequately on its own.
That is not leniency.
It is classification doing the work classification is supposed to do: identifying meaningful differences rather than erasing them.
The preventive lesson is larger than registry policy.
Ask what the allegations require us to notice.
How do trusted adults acquire unsupervised access to children?
How does authority alter the way families interpret concerning behavior?
How does professional prestige change the threshold for suspicion?
When does family loyalty make disclosure harder?
What happens when adults confuse “known to us” with “safe”?
What kinds of boundaries should exist even around people we love?
How can children report something involving a person whom everyone else respects?
Those questions do not offer the emotional simplicity of identifying a dangerous class.
They are harder because they require us to accept something frightening but useful: sexual harm does not reliably announce itself through social identity.
Sometimes the person posing a risk will look exactly like the person we were warned about.
Sometimes he will not.
Sometimes prior convictions will provide relevant warning information.
Sometimes there will be no prior conviction to find.
Sometimes risk enters from outside.
Sometimes it is already at the dinner table.
A prevention system worthy of the word prevention has to be capable of holding all of those realities at once.
Michael Caruso reportedly thought he knew what a sex offender was.
That mental picture apparently excluded himself.
If the factual allegations against him ultimately prove accurate, that intuition failed catastrophically.
But society should resist making the mirror-image mistake after conviction.
Caruso’s reported reasoning was: I cannot be that kind of person, so I could not have done that kind of thing.
Registry culture too easily turns that around: You did a qualifying thing, so now we know what kind of person you are.
Both shortcuts replace inquiry with identity.
Both invite us to stop asking the questions that matter most.
What happened?
How did it happen?
What made access possible?
What does the conduct tell us?
What does it not tell us?
What risk exists now?
What intervention is justified?
What would actually make another victimization less likely?
Sexual harm is real. It can be devastating. Some conduct is extraordinarily grave and warrants profound accountability. Precisely because the stakes are so high, prevention cannot afford mythology, and justice cannot afford categories that masquerade as complete explanations.
Michael Caruso reportedly believed a sex offender could not look like him. Registry policy too often assumes that everyone bearing the label must look like one another. Both ideas substitute an imagined identity for the harder work of understanding conduct, risk, access and human behavior.
If public safety is the goal, we should be much more interested in how harm actually happens than in whether someone looks like the person we were taught to fear.
- Executive Office of the Governor — the arrest affidavit incorporated into Gov. Ron DeSantis’s suspension order. Charges, suspension from office, allegations involving Caruso’s grandson, the fishing-trip and cruise allegations, the recorded February conversation, and statements attributed to Caruso.
- WFLX — WFLX’s reporting on Caruso’s arrest. Contemporaneous local reporting on the arrest and affidavit allegations.
- Executive Office of the Governor — Gov. DeSantis’s 2025 appointment announcement. Caruso’s appointment, Florida House service, and professional background.
- Florida Legislature — Florida’s statutory definition of sexual offender. Statutory classification and registration framework.
- Florida Department of Law Enforcement — Florida’s public sexual-offender and predator system. Public registry search and Florida registration infrastructure.
- Florida Court Clerks & Comptrollers — the Florida clerks’ description of the clerk’s institutional role. Court clerks’ recordkeeping and administrative role.
- Bureau of Justice Statistics — Bureau of Justice Statistics research on sexual assaults of young children. Victim-offender relationships and the prevalence of offenders already known to juvenile victims.
- Bureau of Justice Statistics — BJS’s NIBRS sexual-assault analysis. Sexual assaults recorded by law enforcement and victim-offender-relationship patterns.
- DOJ SMART Office — the Justice Department’s review of sex-offense typologies. Heterogeneity among offending populations and limitations of categorical typologies.
- DOJ SMART Office — the SMART Office’s SOMAPI executive summary. Differences in offending patterns, risk, recidivism, and characteristics.
- DOJ SMART Office — the Justice Department’s review of sexual-reoffense risk assessment. Static and dynamic risk factors, individualized assessment, and intervention targets.
- DOJ SMART Office — the Justice Department’s review of sex-offender management strategies. Registration and notification research, residence restrictions, social versus geographic proximity, and research limitations.