RC Citadel Takedown Desk
policy register verified 2026-07-15
4 providers · routing engine v1

Route a grievance to the layer that can actually act.

Paste a URL or a sitemap. The desk classifies the grievance, screens it for a genuine policy or legal basis, then routes it down the stack — host, registrar, CDN, search — to the party with the authority to act, with a drafted report for each.

1

Intake

single URL, list, or sitemap.xml
2

Classify the grievance

legitimacy gate

Paste the page's text or your notes and the desk ranks the likely basis, showing the signals behind each call. It flags protected content — news, reviews, records — as not a valid basis too. Override anything below.

Run “Identify violation” to rank the likely bases.
3

Resolve the stack

who hosts, registers, fronts the domain
4

Routing plan

ordered venues + evidence
Complete steps 1–3 to generate a routing plan.
5

Drafted reports

tailored per venue
A drafted, policy-citing report is generated for each actionable venue above.

Policy Register

Current abuse channels, accepted categories, and the operational nuance that decides whether a report lands. Each entry links to the canonical source so you can re-verify — providers revise these quietly. See the Field Manual for a Cloudflare Worker that watches these pages for changes.

Field Manual

The reasoning the desk runs on — and the standards that keep reports effective and defensible.

The golden rule of routing

Content removal is a host decision. Only the party storing the files can take a page down. The registrar controls the name, not the content. A CDN in front mostly forwards. A search engine only controls its index.
Domain-level abuse goes to the registrar. Phishing/impersonation domains, look-alike names, and fake WHOIS are the registrar's remit — they can suspend the name itself.
Cloudflare in front means Cloudflare forwards. For pass-through CDN, Cloudflare passes your complaint to the host and gives them the origin IP — useful to reach or unmask the host. It removes content only when it's the actual host (Pages, Workers, R2, Stream, Images) or for phishing/malware/CSAM.
Google delists, it doesn't delete. A successful Google request removes a URL from Search; the page stays live on the web. Pair it with a source-side (host) removal whenever the goal is true removal.

What is genuinely actionable

Providers act quickly on technical and safety violations — phishing, malware, spam/AUP breaches, CSAM, NCII, credible threats — and on rights-based claims you're entitled to bring: copyright in a work you own, your registered trademark used to deceive, exposed personally-identifying data, court-ordered unlawful content. These have clean evidence and clear policy hooks.

What isn't — and why forcing it backfires

Accurate journalism, honest reviews and opinion, public and court records, and true-but-unflattering information are not policy violations. Abuse desks are built to reject these, and pushing them through the wrong door has real costs:

Costs of over-claiming
  • A DMCA notice on material you don't own is a sworn statement made under penalty of perjury — bogus notices expose you to liability under 17 U.S.C. § 512(f).
  • Repeated invalid reports get a sender de-prioritised or blacklisted by an abuse desk, weakening your legitimate filings.
  • Takedown attempts on lawful content routinely trigger the Streisand effect and counter-coverage.

When content is lawful, the honest routes are: a factual-correction or right-of-reply request to the publisher; narrow GDPR/"right to be forgotten" delisting where an EU data subject qualifies; suppression through owned assets you control; or — if it is actually defamatory — a legal process that ends in a court order, which then unlocks the court-order removal path at Google and the host.

Evidence standards

Every report is only as good as its proof. Bring: the exact offending URL (never just the homepage); a timestamped screenshot; for email abuse, full plain-text headers; for malware, a VirusTotal or sandbox link; for rights claims, the registration number / link to your original work and a statement of authorisation. De-fang live malicious URLs (hxxps://, [.]) so filters don't strip your report.

Continuous policy monitoring (Part A, wired for real)

A static page can't poll live policies, so pair this desk with a scheduled watcher. The companion file policy-monitor-worker.js is a Cloudflare Worker (cron-triggered) that fetches each canonical policy URL in the register, hashes the content into Workers KV, and emails you via Resend when a page changes — so "verified" is a live date, not a snapshot. Deploy notes are in the handoff.

Scope

This desk organises reporting workflow and drafts correspondence. It doesn't determine whether content is unlawful — that's a legal judgment. For defamation, court orders, or cross-border questions, route through counsel (Taylor Wessing / Farrer & Co. on the referral list).

copied