Legal architecture
The riskiest sentence in the concept
"I don't care if they were grandfathered in." It is the one design choice most likely to end the program in court - and the one with the cleanest fix.
Courts have tolerated retroactive registration for over twenty years. They keep drawing the line at retroactive monitoring and geographic restriction. The distinction is not a technicality; it is the whole doctrine, and a design that respects it can do almost everything the concept wants.
In 2003 the Supreme Court upheld retroactive registration and notification as civil rather than punitive. That case is the reason people assume retroactivity is settled. It is not, because monitoring is not registration.
Smith v. Doe, 538 U.S. 84 (2003)
In 2016 the Sixth Circuit held Michigan's retroactive registry amendments punitive, and so an ex post facto violation. The exclusion zones were the specific feature the court identified as punitive. In 2017 Pennsylvania's Supreme Court reached the same conclusion about its own registration statute.
Does #1-5 v. Snyder, 834 F.3d 696 (6th Cir. 2016), cert. denied, 138 S. Ct. 55 (2017)
Commonwealth v. Muniz, 164 A.3d 1189 (Pa. 2017)
And in 2015 the Supreme Court held that attaching a device to a person to track their movements is a Fourth Amendment search. That case decides something different but equally structural: a monitoring program cannot describe itself as merely administrative. It is a search, and searches need justification proportionate to the individual searched.
Grady v. North Carolina, 575 U.S. 306 (2015) (per curiam)
The constitutional map
Which of these is settled and which is a live split decides where a program can be deployed and where it cannot, so the status column is the one to read first. The already law page gives the four cases that most directly shape this design in plain English; this is the fuller authority map behind them.
| Case | Holding | Status |
|---|---|---|
| Grady v. North Carolina [6] | Attaching a device to a person's body to track their movements is a Fourth Amendment search. Reasonableness expressly not decided. | Settled |
| Smith v. Doe [7] | Registration alone is civil and may apply retroactively - reasoning resting on the absence of physical restraint or supervision. | Settled |
| Packingham v. North Carolina | Blanket social-media bans on registrants violate the First Amendment. Narrowly tailored conduct restrictions preserved. | Settled |
| Chatrie v. United States [9] | A reasonable expectation of privacy in cell-phone location information, with no durational floor. Brevity is no longer an answer. Geofence-warrant validity and the third-party doctrine are expressly left open. | Settled in part |
| Does #1-5 v. Snyder [8] | Retroactive exclusion zones, plus burdensome reporting, plus offense-only tiering, amount to punishment: an ex post facto violation. | Circuit split |
| NARSOL v. Stein [10] | Upheld retroactive 1,000-foot residency restrictions against an ex post facto challenge - substantially what the Sixth Circuit struck down. | Circuit split |
| Ellingburg v. United States | Restitution is criminal punishment for ex post facto purposes - decided on text and structure without reaching the Smith multifactor test. | New method |
The split is genuine, and a national program built on retroactive imposition is betting on which circuit it gets sued in
Two federal circuits reached opposite conclusions about substantially the same feature, and no higher court has resolved it. A design that cites only the half it likes will be taken apart by the first person who reads the other half. The workable substitute is not weaker, it is slower: impose prospectively at sentencing, and give people already in the system an individualized hearing path. That reaches the same population over time without the facial challenge.
Why Ellingburg deserves attention from anyone drafting
The Court resolved a punishment question through textual and structural analysis - where the obligation is codified, who imposes it, at what proceeding - and once text and structure showed punitive intent, the inquiry ended. The Smith multifactor test was never reached, which is a different thing from being declined: the Court did not weigh those factors and reject them, it never arrived at them. It does not overrule Smith v. Doe, and no court has yet applied it to monitoring. But it signals that formal placement now carries independent weight, and that cuts directly against imposing monitoring at sentencing in a criminal code while intending to defend the scheme as civil. Whether any court has applied that method to a registration or monitoring scheme is one of the items still open.
The supervision line
This is the sharpest distinction in the area, it is load-bearing for who can be enrolled at all, and it is not settled.
For a person still under supervision - probation, parole, post-release supervision - monitoring rides on the diminished expectation of privacy arising from an ongoing custodial relationship. No warrant, no individualized suspicion. Nationally this is close to settled in the government's favor.
For a person who has fully discharged their sentence, that rationale evaporates by its own terms, and courts have split squarely:
| Court | Case | Result |
|---|---|---|
| Seventh Circuit | Belleau v. Wall, 811 F.3d 929 (2016); Braam v. Carr, 37 F.4th 1269 (2022) | Upheld No individualized suspicion required. The court leaned on monitoring that tracked location only, with no real-time review and no movement restrictions. |
| Georgia | Park v. State, 825 S.E.2d 147 (2019) | Struck facially Post-release individuals are "free persons"; the special-needs rationale was rejected as not divorced from general law enforcement. |
| North Carolina | State v. Grady (2019); Hilton, Strudwick (2021) | Split by category Struck for unsupervised recidivists; upheld for aggravated offenders. |
| Massachusetts | Commonwealth v. Arnold, SJC-13768 (2025) | Duration-limited Monitoring permitted only for a duration the judge finds reasonable, even if shorter than the probationary term. |
Design decision D-4a - scope the initial deployment to people under active supervision
It is the only population where the Fourth Amendment question is close to settled. It avoids the entire post-discharge split. And it is where a supervising agency - the actual buyer - already holds both the authority and the budget. Post-discharge extension is a phase-two question that should follow a Supreme Court answer rather than precede one.
Note what the Seventh Circuit relied on in upholding: no real-time review, and no movement restrictions. This design has real-time detection at its center, which means the Belleau reasoning does not transfer to it unexamined. That is a reason to enrol inside the supervision relationship rather than outside it, and it is worked through in due process.
Why blanket retroactive imposition fails
"I don't care if they were grandfathered in" is the single riskiest premise available here. Three lines of authority make it the most likely point of failure.
- Smith rested on the absence of the very thing this design adds. Retroactive registration was permitted precisely because Alaska's scheme involved no physical component - registrants remained free to move, live and work as other citizens, with no supervision. A scheme that attaches a device and restricts movement removes the exact features the holding rested on.
- Courts keep naming the same three features as punitive. Geographic exclusion zones; burdensome in-person reporting; and tiering based on the offense of conviction with no individualized assessment. That triad drove Does v. Snyder and Doe v. Lee, and a number of state high courts have reached similar conclusions on state constitutional grounds - Alaska, Indiana, Maine, Ohio, Maryland, Oklahoma, New Hampshire and Michigan. An earlier version of this page counted them instead of naming them; a count cannot be checked, and a list can.
- The split is genuine and unresolved. A national program built on the assumption that retroactive imposition is lawful is betting on which circuit it gets sued in.
An appeal directly material to this question was pending when the underlying research was done, and its outcome is unknown. It is carried as an open item rather than resolved here - see open questions, and check it against primary sources before anything is filed or funded.
Tokens are not authority
The single most consequential error available here, and it is an attractive one
The argument runs: because location data is tokenized and no standing database exists, tracking is not legally constrained. It is wrong, it is wrong in a way that is expensive to discover late, and correcting it now costs nothing.
Grady locates the search at the attachment of the device to a person's body and the tracking of their movements [6]. The search is therefore complete before any datum is stored, encrypted, hashed, or discarded. What happens to the resulting data afterwards cannot determine whether a search occurred, because the search already occurred. Data handling goes to reasonableness - and it goes there as one factor among several, not as an answer.
Chatrie tightens this [9]. It holds that an individual has a reasonable expectation of privacy in cell-phone location information, and it removes the durational threshold that had been left open - highly sensitive personal information can be gleaned from even small periods of location history. Arguments that a collection is de minimiz because it is brief, coarse, or anonymized are weaker now than they were when this area of doctrine was last widely written about.
| What tokenization does | What tokenization does not do |
|---|---|
| Materially strengthens the reasonableness side of the balance. A system holding no queryable movement history cannot be put to the general-law-enforcement purposes that sank a state statute in Georgia; it cannot be browsed by an insider; it cannot be breached in bulk; it narrows the intrusion to the specific event that triggered it. The Seventh Circuit leaned explicitly on monitoring that involved no real-time review and no movement restrictions. Architectural minimalism is a real legal asset, and it should be argued for as one. | Affect whether a search occurred. It does not, and no amount of it will. It also does not deliver anonymity in this threat model, because the adversary of concern can physically observe the person whose identifier they captured, and no cryptography defends against an adversary with eyes [28]. |
The ballot analogy that motivates this design is close, and it breaks at one specific point. A ballot can be perfectly anonymous because nobody ever needs to act on an individual ballot. This system must send an officer to a specific person at a specific place, so there must be some path from identifier to identity. The honest design is therefore not "no database". It is no standing database, and a controlled, logged, judicially authorized resolution path - which is what the data flow specifies and what the identity page governs.
Say this plainly in any filing or pitch, because the alternative is that opposing counsel says it first. Tokenization is a strong argument on reasonableness. It is not a reason the Fourth Amendment does not apply.
The test that decides it
Both Snyder and Muniz reached their conclusions by applying the seven factors are used to decide whether a nominally civil scheme is punitive in effect. A scheme the legislature calls civil is treated as punishment - and therefore cannot apply retroactively - if its effects are punitive under these:
Kennedy v. Mendoza-Martinez, 372 U.S. 144, 168-69 (1963)
| # | Factor | How a proximity-monitoring program scores |
|---|---|---|
| 1 | Affirmative disability or restraint | Adverse A device on the body plus exclusion zones is a restraint on its face. This is the factor Snyder turned on. |
| 2 | Historically regarded as punishment | Adverse Supervision and banishment from places both have direct historical analogues. A public audible alarm adds shaming, which has the clearest analogue of all. |
| 3 | Comes into play only on a finding of scienter | Adverse if imposed by administrative reclassification. Neutral if imposed at sentencing on the conviction itself. |
| 4 | Promotes the traditional aims of punishment | Mixed Deterrence is an explicit aim of every such program, and deterrence is a traditional aim of punishment. |
| 5 | Applies to behavior already criminal | Adverse Unavoidably. Enrollment is triggered by a conviction. |
| 6 | Rationally connected to a non-punitive purpose | Favorable Public safety, and this design can show its work: a tier ladder, an evidence base, and a published coverage model. |
| 7 | Excessive in relation to that purpose | Decisive This is where the design is actually won or lost. Lifetime, blanket, no hearing, no exit is excessive. Term-limited, individually assessed, reviewable, with a defined exit, is not. |
The architecture that survives
- Imposed prospectively at sentencing, as an express component of the sentence, on the record, with the defendant on notice before plea.
- Existing registrants only through an individualized hearing - a validated risk assessment, notice, the right to be heard, the right to counsel, and a written finding. Not a blanket sweep.
- A defined term tied to the assessment, with mandatory periodic review rather than lifetime by default.
- A petition for removal that can actually succeed, with published criteria and a real grant rate.
- Exclusion zones drawn narrowly, if at all - they are the specific feature that lost in Snyder, and they are the least useful part of the design once the child-side device exists.
Score your own design
Nine choices, and what each one costs in court
Every switch below corresponds to a decision a drafting committee will actually have to make.
Retroactivity and punitive-effect risk scorer
Each choice cites the authority it is drawn from-
The weights are a drafting aid, not a prediction, and no scorer substitutes for counsel in the relevant jurisdiction. They are ordered by how directly each choice tracks a holding: retroactivity and exclusion zones carry the most weight because they are the two features Snyder named. The law here is genuinely split, and both halves of it are recent. In 2016 the Seventh Circuit upheld lifetime GPS monitoring against Fourth Amendment and ex post facto challenges. And in 2019 North Carolina's Supreme Court held lifetime satellite-based monitoring unreasonable as applied to a recidivist no longer under supervision. A design that only cites the cases it likes will be taken apart by the first person who reads the rest.
Belleau v. Wall, 811 F.3d 929 (7th Cir. 2016)
State v. Grady, 372 N.C. 509, 831 S.E.2d 542 (2019)
What the statute has to say
These are drafting requirements, not principles. Each one exists because a court has struck down the alternative, and each one is specific enough to hand to legislative counsel.
| Requirement | Why | |
|---|---|---|
| L-1 | Prospective imposition only - ordered at sentencing, or at a subsequent individualized hearing. No blanket retroactive sweep. | The retroactivity split above. This is the substitution that yields the same population over time without the facial challenge. |
| L-2 | An individualized judicial finding: a validated risk assessment, plus an explicit finding that this person requires this level of monitoring. | Category-based automatic imposition is the single most frequently invalidated feature in the case law. North Carolina's 2021 model, S.L. 2021-138, is the template. |
| L-3 | A definite duration with periodic review. North Carolina replaced lifetime monitoring with a fixed 10-year term and a petition for termination after 5 years. | Arnold goes further: duration is itself part of the reasonableness inquiry and has to be justified across the whole ordered period, not just at the start. |
| L-4 | Risk-tiered, not conviction-tiered. | 0.7% to 48.7% is the range of five-year sexual recidivism across Static-99R scores - roughly a seventyfold spread.[26] Treating that population as uniform is empirically wrong and is the exact defect courts identify as punitive. |
| L-5 | Narrow the covered class: define by child-victim offense of conviction plus risk tier. | "Paedophile" is not a legal category, and registries include offenses with no child victim and no sexual component at all - see due process. |
| L-6 | No public identity disclosure from an alert. An alert must not name, display or describe the monitored individual to any member of the public. | A loud public alarm de-anonymizes instantly, and registered individuals have been killed using registry data. This is why the audible tier in the alert tiers sits behind three prior gates and a human. |
| L-7 | Judicial authorization for identity resolution, with a narrowly defined exigency exception, and an immutable audit record on every resolution. | The whole architecture of the data flow depends on this being a legal control and not only a technical one. An exigency path with no time limit and no review defeats the architecture by its own exception. |
Hanson, Thornton, Helmus & Babchishin (2016), Sexual Abuse 28(3); Static-99R Evaluators' Workbook (2021)
Treating that population as uniform is both empirically wrong and the exact defect courts identify as punitive.
Decide early whether the scheme is civil or criminal, then stay consistent
The two paths are mutually exclusive in a way that is easy to miss until it is expensive. A scheme defended as civil and regulatory survives ex post facto challenge, but it argues itself out of Eighth Amendment protection and cannot be described as part of the sentence. A scheme that is punishment gets Eighth Amendment proportionality review and cannot be applied retroactively. There is no version that takes the advantages of both.
Post-Ellingburg, where the obligation is codified and who imposes it may settle this question before anyone gets to argue it.
Implanted devices, and why abandoning them was right twice over
The devices page refuses implants on technical grounds: a passive implant has no battery, no satellite receiver and no radio of its own, is readable only at inches, and would be strictly worse at this job than a wearable. That half of the reasoning stands on its own. This is the other half.
There is no United States case law directly on mandatory implanted tracking devices, so the analysis is by framework rather than by holding - and the framework is unpromising. Winston v. Lee held compelled surgery an unreasonable search. Birchfield v. North Dakota drew its line precisely at piercing the skin. Washington v. Harper, 494 U.S. 210 (1990), establishes a significant liberty interest against unwanted bodily intrusion. Skinner v. Oklahoma, 316 U.S. 535 (1942) - decided on equal protection grounds rather than substantive due process - shows how skeptically courts treat irreversible bodily alteration of offenders. Layered on Grady and Chatrie, a mandatory subdermal tracker would face a very high probability of invalidation.
An earlier version of this page cited both cases together as substantive-due-process authority on bodily integrity. That is a fair summary of where the two point and a poor description of what either held, and the ground a case was decided on is the first thing an opponent checks. The citations above are the ones to use.
Permanent physical marking has no plausible constitutional defense at all, and none is offered here.
The other half nobody litigates yet
All of the case law above concerns the monitored adult. The child's band raises separate questions that have no settled answer, and a program that has not thought about them will meet them in a complaint:
- Who consents? A guardian can consent for a young child. A fifteen-year-old's own view is legally relevant in a growing number of contexts, and a device they did not agree to is a device they will disable.
- Schools. Wearables policy, and whether a school can require or forbid the band. Both happen.
- Custody disputes. A location device in a contested custody arrangement is an instrument of that dispute. The design's answer - no retrievable track, ever - is not incidental; it is what keeps the band from being subpoenaed into a custody fight.
- Children in state care. Who is the guardian for enrollment purposes, and what happens at placement changes.
- The anchor log as evidence. Anything an anchor records is discoverable, including the Tier 0 entries that were never transmitted anywhere. The 72-hour rolling window is a retention decision with an evidentiary consequence - for the monitored person and for anyone else who walked past - and it should be made deliberately by counsel rather than by an engineer choosing a buffer size.
How this works
0:58 · narrated · captioned · nothing here is sent anywhere when you press play. Prefer to read it? The transcript is below.
Transcript
| At | Section | What it says |
|---|---|---|
| 0:02 | Where it breaks | Courts have tolerated retroactive registration for over twenty years. They keep drawing the line at retroactive monitoring and geographic restriction. |
| 0:11 | The split is genuine | Two federal circuits reached opposite conclusions on the same feature. A program built on retroactive imposition is betting on which circuit it is sued in. |
| 0:21 | Tokens are not authority | Under Grady the search is complete when a device is attached and tracks, before anything is stored or hashed. Tokenization is a privacy control, not a legal authorization. |
| 0:33 | What the design does | So enrollment is prospective at sentencing. Existing registrants come in only by individualised hearing, the term is definite, and a petition for removal can actually succeed. |
| 0:44 | Score your own design | The page carries a scorer you can run against a proposed design. Nine choices, weighted by how directly each one tracks a holding. It is a drafting aid, not a prediction. |