# D001 — Antithesis: The Topology Is Product Design, Not Policy Failure ## Counterclaim The 1,384-org Conway PD Flock sharing topology is a structural property of the Flock product, not the residue of governance discipline that failed to deploy. Flock's own admin interface treats some form of sharing as the operational baseline — the configuration menu offers five sharing modalities of which "don't share at all" is merely the fifth and last option, and the path of least resistance through that menu produces wide sharing rather than narrow sharing. Flock has demonstrated, in its Virginia and Illinois deployments, that it possesses the structural capacity to gate sharing at the *product* level when law compels it — and has declined to do so anywhere law does not compel it, including Arkansas. CPD 800-32 §D.4's "evidence of an offense is indicated" standard was drafted in the pre-cloud era for an architecture in which "sharing" meant an officer's individual release of a specific record; it is a per-event evidentiary test, and it has no operative handle on a platform whose unit of action is a standing relationship toggle that authorizes thousands of subsequent queries the configuring administrator will never see. The thesis's per-event policy frame is a category mismatch. The regulated object — the entity whose design choices produce the topology — is the vendor. The remedy is structural restriction at the platform level, of the kind Virginia and Illinois already extracted from Flock, not exhortation toward configuration discipline against the very interface affordances that make that discipline structurally unavailable. ## Attack on the thesis The thesis's `## Argument` section opens with a textual sleight: it asserts that CPD 800-32 §D.4's "evidence of an offense is indicated" language is "general enough — and forceful enough — to attach to the moment of platform-level configuration." This is wishful exegesis. §D.4 reads "LPR Data captured by the department's LPRs **may be shared** with other law enforcement agencies **if evidence of an offense is indicated**." The unit of analysis is *data* and the act regulated is *sharing the data*. The provision presupposes a specific quantum of data being released and asks whether evidence of *an offense* — singular, definite, identifiable — is indicated. That is the grammar of a per-event evidentiary test, indistinguishable in form from the Fourth Amendment particularity requirements the drafters were almost certainly thinking with when they wrote it. Stretching that text to govern the standing-relationship configuration moment requires reading the singular "an offense" as if it meant "the abstract possibility that evidence of some unspecified offense might in the future come to be indicated by some query of some hot list maintained by the counterparty agency on some plate that the counterparty agency has not yet identified." That is not what the policy text says, and the thesis's own `## Anticipated counterarguments` section concedes the point — it lists the "literal-reading objection" first and notes that "the policy text references 'the department's computer servers'" and that reading "share" as reaching standing inter-agency network relationships "imports a meaning the drafters could not have anticipated." The thesis acknowledges the objection and declines to refute it. The acknowledgment is correct. The decline is fatal. A policy text that the policy's own defenders concede was written for a per-event physical-camera architecture cannot do the work the thesis demands of it. The thesis's `## Argument` then leans on what it calls the **human-decision** premise: "The configuration was a human decision, performed in a platform admin interface, by an identifiable Conway PD administrator." This is true and irrelevant. Every default state in every piece of software was configured by some human at some point — by Flock's product team, choosing the resting state of the sharing toggle when an agency onboards. The thesis's move is to locate the "human decision" exclusively at the agency end, as if the absence of an administrator's affirmative override constitutes that administrator's "choice." This is the same fallacy that makes terms-of-service click-throughs into binding contracts: it equates the failure to reverse a vendor-engineered default with positive agency consent to it. The thesis concedes the move with the throwaway line "either way, the configuration is the agency's" — a sentence that does the work of erasing the design choice the vendor made upstream of any toggle the administrator did or did not flip. That erasure is the analytic load-bearing point on which the thesis depends; it does not survive examination. The thesis's `## Argument` section then cites the Flock SVP's "you are responsible for knowing your agency's laws and policies" line as a vendor *concession* of policy-neutrality. This badly misreads the Flock document. The full sentence — Tier-1 verbatim, see Evidence below — is not a concession; it is a *disclaimer*. It comes immediately after Flock has just described its own product-level restrictions in Virginia and Illinois ("**in some states, sharing is automatically restricted as required by law**"). The structure of the paragraph is: *we build structural restrictions where law mandates them, and where it does not, you are on your own.* That is not the vendor saying "the agency holds the governance posture." It is the vendor saying "we have designed the product to do what the law of your jurisdiction makes us do, and otherwise we have designed it to share — and you are on your own to talk your way back from that default." Citing this disclaimer as evidence of vendor policy-neutrality is the analytic equivalent of citing a tobacco company's surgeon-general label as evidence the company believes its product is safe. The thesis's `## Argument` then deploys the asymmetric-configuration evidence — 471 inbound-only, 427 outbound-only — and argues it "does not rescue the product-design reading; it indicts the configuration-discipline reading further." This gets the inference exactly backward. If the asymmetries had been the product of a per-relationship application of §D.4, the asymmetries would be *justifiable*: a per-event analysis of Counterparty A might warrant inbound-only sharing because Counterparty A's hot lists are valuable to Conway operations while Counterparty A's operational profile does not present "evidence of an offense" justifying access to Conway's data, and so on across hundreds of counterparties with case-by-case warrants. No such case-by-case analysis is documented anywhere in the corpus. The asymmetries are unexplained — and the thesis's own concession that they were either "deliberate" or "inherited from defaults the administrator never reviewed" is a binary that hides the obvious third possibility the corpus actually supports: that the asymmetries reflect counterparty-side toggle choices Conway's administrator had no part in. A relationship is bidirectional if *both* parties' administrators have toggled on outbound sharing; if only one has, the relationship is one-directional. 471 of the inbound-only relationships and 427 of the outbound-only relationships are therefore not principally evidence of *Conway's* discipline or its absence — they are evidence that other agencies' administrators, deploying the same product under the same default-on UX, made independent toggle decisions that happened not to mirror Conway's. The asymmetric topology is therefore a *structural property of a network of independently configuring default-on tenants*, not of any one tenant's configuration discipline. That is a product-design fact, not an agency-discipline fact. The thesis's `## Evidence` section produces no document that shows Conway's administrators actually applied §D.4's standard to any of the 1,384 configuration decisions — no approval logs, no per-relationship evidentiary memoranda, no chain-of-command sign-offs. The thesis acknowledges this absence and recasts it as evidence of agency negligence. The same absence is equally consistent with the simpler hypothesis that no such per-relationship application is *possible* in the product interface, because the product was not built to ask the question. The thesis's framing requires the reader to assume that the absent evidence is absent because it was never created when it should have been; the antithesis's framing observes that the absent evidence is absent because the product affords no slot in which to create it. Without independent evidence selecting between these hypotheses, the policy-failure account is doing the heavy lifting of the inference unaided. ## Independent argument for the counterclaim The dispositive fact, available Tier-1 in the Aug 2025 Flock explainer, is that *Flock itself has built structural sharing restrictions where law has compelled them and has declined to build them where law has not*. The vendor's own SVP, in his own words to Conway PD's Flock administrator, writes: "**in some states, sharing is automatically restricted as required by law, and searches with search terms that indicate a purpose prohibited by law have been disabled in our product. For example, in Virginia, out-of-state sharing is disabled and in Illinois, accessing data for certain purposes is not allowed.**" This sentence is the entire ball game. Virginia and Illinois did not extract their structural restrictions from Flock by Virginia and Illinois police agencies' configuration discipline. They extracted them by *statute*, and Flock implemented them at the *product* level. The capacity to gate sharing at the product level is therefore *demonstrably available* to Flock. Where the capacity exists and the vendor declines to deploy it, the default-on posture is a positive design choice — not an artifact of agency discipline that failed. Arkansas has no statute analogous to Virginia's out-of-state-sharing prohibition. The Arkansas-deployed product therefore ships with the default-on posture that produces a 1,384-org topology, and that posture is what the product *is*, not what the agency made of it. To argue, as the thesis does, that Conway PD could have produced a 1,384-relationship pruning analysis matching §D.4 to each counterparty is to argue that the agency could have manually replicated, at the configuration end, the structural gating that the vendor declined to build at the product end. That is a structurally backwards demand. The vendor builds the cage or the vendor does not; agencies do not retro-fit cages around their own daily operations. The Flock product's menu of sharing options confirms the design reading. Flock's own SVP enumerates five options: "**Share with agencies in specific states only · Share with agencies with similar laws · Share within your state only or within a certain distance · Share 1:1 · Don't share at all.**" Four of those five options are forms of sharing. The fifth is the option not to. The menu's typography and its presentation order matter — "Don't share at all" is the terminal option, an after-the-fact reservation rather than an architectural baseline. A product whose architectural baseline is non-sharing would invert the menu: it would default to no sharing and present *opting into* sharing as the configuration choice. Flock's menu does the opposite. The administrator who does not actively narrow the configuration is therefore not in a state of non-sharing; she is in some state of sharing whose specific scope was determined by the vendor's default for that customer, in that jurisdiction, at that contract tier, at that point in the platform's release history. None of those determinants is the agency's, and none is governed by §D.4. They are all the product's. CPD 800-32 was drafted in 2013 for a per-event, physical-camera architecture that the Flock platform has obsoleted. The thesis itself notes that "the policy text references 'the department's computer servers'" — i.e., a world in which LPR data lived on department-owned hardware and "sharing" meant the affirmative release of a specific record from that hardware to a specific outside agency. That world is gone. The current architecture stores LPR data in Flock's cloud, makes it queryable by any organization with whom a sharing toggle has been flipped, and produces no per-query agency awareness on Conway's end. The "evidence of an offense" standard presupposes that the agency knows a sharing act is occurring and can therefore evaluate the evidence at the moment of occurrence. The contemporary product makes Conway invisible to most of its own sharing events: the [[Federal Searches CSV]] documents 5,929 federal-LE queries against Conway's data in a 51-day window, none of which Conway approved at query time, most of which Conway had no contemporaneous notice of. A per-event evidentiary policy cannot govern a system in which the policy-applying agency does not know per-event when the events are happening. The policy text the thesis stakes its case on is doctrinally obsolete on its own terms — and the source of that obsolescence is the product architecture that has separated the act of sharing from the agency's per-event awareness of it. The asymmetric configurations and the enabled-but-unused federal relationships are the signature of default-on configuration, not of per-event evidentiary analysis. The corpus documents `[Federal] Wright Patterson OH Air Force Base` as one of the 1,384 organizations — a relationship that *exists* but that produces no recorded activity in the Federal Searches CSV's available window. A per-event policy test asking "is evidence of an offense indicated" cannot have authorized a relationship that has never been used to indicate evidence of any offense at all. The relationship exists not because Conway's administrators looked at it, asked the §D.4 question, and answered yes; it exists because the product made it exist, by default, when some upstream toggle — possibly counterparty-side, possibly Flock-side, possibly an artifact of the federal-prefix convention itself — placed Wright Patterson into the network of Conway-reachable organizations. The same logic applies to the 898 one-directional relationships. The structure is the structure of a *network of independently configuring default-on tenants*, not the structure of a per-relationship evidentiary review. The thesis insists on reading the topology as the residue of agency choice; the topology reads instead as the emergent product of a vendor design that has made the choice unreviewable. The remedy follows. If the topology is a product property, then directives from Chief Harris cannot reach it — they can only exhort administrators to do, at the configuration end, what the vendor's design has made structurally unavailable. The remedy must operate where Virginia's and Illinois's remedies operated: at the platform level, by statute, compelling Flock to design the product such that the default state is no sharing and sharing requires affirmative per-relationship justification entered into the product itself. That is product-design regulation, and it is the only intervention that can produce the structural restriction the per-event policy text was originally designed to enforce. ## Evidence **The Flock vendor's product-level structural restrictions, verbatim Tier-1**, from the 2025-08-20 explainer (Chris Colwell, SVP of Customer Experience, Flock Safety, to Lt. Andrew Burningham; raw at `D:\arkansas-surveillance\extracted\conway-pd\PD-2026-354\Flock Safety Understanding Sharing Search and Audits in the Flock LPR System.txt`, lines 25–29): > "In some states, sharing is automatically restricted as required by law, and searches with search terms that indicate a purpose prohibited by law have been disabled in our product. For example, in Virginia, out-of-state sharing is disabled and in Illinois, accessing data for certain purposes is not allowed. > > However, you are responsible for knowing your agency's laws and policies, and ensuring that you and your agency's users are using the Flock system in compliance with these rules." This is the load-bearing sentence. Flock builds product-level restrictions *where law compels them*; otherwise Flock disclaims responsibility and the default-on product ships unmodified. **The five-option sharing menu, verbatim Tier-1**, same source, lines 22–24: > "National sharing — Opt into Flock's national sharing network… Share with agencies in specific states only · Share with agencies with similar laws (for example, regarding immigration enforcement and data) · Share within your state only or within a certain distance · Share 1:1 — Share only with specific agencies you have selected · Don't share at all." Four of the five options are forms of sharing; the option not to is the fifth and last. The menu's architecture treats sharing as the baseline configuration mode. **The data-ownership / system-administrator framing**, same source, lines 16–17, which frames the entire configuration responsibility as falling on the customer: > "Each LPR camera in your network captures vehicular evidence to help solve crime. That data belongs to you as the customer. As the system administrator, you have full control over whether and how your agency shares this information with others." Read alongside the structural-restrictions paragraph, this is the vendor explicitly displacing onto the agency a configuration burden the vendor's product architecture has made structurally impossible to discharge at the per-event level §D.4 contemplates. **The asymmetric structural facts** from [[SharedNetworks 2025-12-17 Snapshot]]: 1,384 total organizations; 486 bidirectional; **471 inbound-only**; **427 outbound-only**; 94% interstate; the federal entity `[Federal] Wright Patterson OH Air Force Base` present with no recorded activity in the [[Federal Searches CSV]]'s available window. These are the topology fingerprints of a default-on configuration model with independently toggling counterparty tenants — not the topology of a per-relationship evidentiary review applied 1,384 times. **CPD 800-32's per-event grammar**, anchored at [[CPD Policy 800-32 — License Plate Reader Vehicle Operations]] §D.4 and quoted in the [[D001 Thesis]] Evidence section: "LPR Data captured by the department's LPRs may be shared with other law enforcement agencies if evidence of an offense is indicated." The grammatical unit is *data*; the regulated act is *sharing the data*; the trigger is *evidence of an offense* — singular, identifiable, present. This is per-event evidentiary policy grammar. It has no operative handle on a per-relationship product whose unit of action is a one-time toggle.