9 September 2026. Root acquisition is closed. Four original company HTML pages were retained and their full substantive text read. A current SEC-hosted FY2025 filing was read in specific passages through the primary web reader; no original filing was downloaded. These are company descriptions, not a customer configuration, independent accuracy audit or observed restoration. The historical FilmOn case is a separate lane. Current brand/product statements and consolidated issuer disclosures do not by themselves prove the exact contracting entity in the older lawsuit.
Category assignment and customer selection
The undated current [brand-suitability product page](http[local research file] describes a proprietary classifier using AI and human review across several media types. It supplies more than100 categories which brands can choose to avoid. Its ordinary commercial example includes avoiding otherwise legitimate video-game or motor content to fit a particular campaign. The page distinguishes pre-bid evaluation from post-purchase measurement. Neither classification nor measurement alone demonstrates a completed advertising exclusion. The advertised accreditation and accuracy remain vendor claims here; no accreditor audit or actual campaign was inspected. Original-derived lines59–111, particularly78–98; primary reader77–184.
An announced default complicates the choice model
The [13 February2025 announcement](http[local research file] announces 'Highly Illicit: Do Not Monetize' and 'P2P Sharing and Streaming' avoidance categories and says both will be enabled by default. This establishes an announced default policy; the activation date and customer uptake are unverified. It reports automatic provision of the former to advertisers and distribution to over100 platform partners. A supplied category or default does not prove every recipient enabled it or every listed site's advertising stopped. The latter covers infrastructure that the text acknowledges has legitimate uses but can be misused.
DV says it used three years of public NCMEC data, covering over300 domains/providers, cross-checked prior classifications and assessed remaining sites itself. It says it excluded large, well-moderated social/search services. This is DV's account of input selection and transformation, not an independently verified NCMEC judgment or endorsement of DV's resulting list. The underlying NCMEC table and its meaning were not acquired in this run. Named FBI outreach is an offered-assistance account, not acceptance or instructions. Original-derived72–138, decisive96–120; primary reader92–139. The captured original supplies the publication date.
Review and evidence access
A [13 June2024 company statement](http[local research file] describes a publisher classification-review process, access to reports/logs, and separation of assessment from sales/client teams. It claims paying-publisher status does not soften ratings and gives an unnamed customer example. These are safeguards asserted by the company, not independently observed separations or a worked correction. It supplies no decision deadline, independent appeal adjudicator, mandatory advertiser reconsideration or restored-income outcome in the read body. Its automation/objectivity language does not establish absence of model or policy bias. Original-derived72–100; primary reader94–115.
The [25 February2025 URL-reporting announcement](http[local research file] says DV advertisers using Brand Safety and Suitability measurement can request URL-level detail. Direct integration into its interface was 'coming soon' and external integrations were planned later that year. Those are stages in a dated announcement; retrieval in2026 does not turn the plans into completed deployment. The described entitlement is for advertiser customers and does not itself give every affected publisher access to every customer's report or preferences. No request was made. Original-derived72–97; primary reader95–118. Actual dates on the captured originals resolve the web-reader's omitted publication dates; nearby recent-post dates are not used.
The revenue model and configuration authority
The [FY2025 Form10-K](http[local research file] describes customer configuration distributed across the service, proprietary classification and customer feedback shaping product development. It reports90percent of2025 revenue from advertisers and10percent from supply-side customers. Advertiser fees are primarily based on measured transaction volume, commonly per thousand impressions; some percentage-of-spend arrangements are capped. Supply-side arrangements include minimum guarantees and tiered pricing. This supports a two-sided commercial model, not a fee per adverse classification or a proved financial cause of a specific exclusion.
Exact retained source scope is product-10k.reader.txt: reader403–434 (customer feedback, product teams and configuration/classification) and1152–1225 (MD&A overview, fee basis and revenue sources). The relevant prose was read; no full filing, audit opinion, revenue-table facsimile, individual contract or accounting record was inspected. The fiscal year is2025, not the capture year. No stockholder command, grant finance or external ownership branch was acquired.
Structural result and next discriminator
The current offering contains several powers that should occupy separate places on the map: selecting source inputs, assigning a classification, choosing a default, changing customer settings and implementing a placement decision. A review can change one without necessarily changing the others. The acquired product description supports customer choice; the dated announcement supplies a vendor-chosen default policy, with implementation not independently observed. Neither cancels the other.
The most revealing next input record is the exact public NCMEC dataset DV says it used, including whether its rows count provider reports, notices received or assessed illegal domains. Those meanings are not interchangeable. The missing implementation records are the category policy/version, a recipient's actual configuration and a dated review/correction followed by recipient action. The historical court lane may illuminate the contested classification process but cannot prove these2025/2026 settings were used then.
No customer login, publisher submission, operational classification query, harmful-content endpoint, payment, outreach or public release occurred. Acquisition closes pending independent review and the combined forest. All original files and derivative scopes are listed in product-captures.json; successful captures are not claims that every link or surrounding page was read.