Powered by AI · Executed by humans

Take it down.
And keep it down.

Built for fast leak triage, evidence-backed web notices, and reviewer-led social platform work. The system keeps the automation narrow and the hard calls human.

The model

Powered by AI. Executed by humans.

Speed comes from the model. Judgment comes from the team. Every finding gets both — within minutes, not days.

AI does what AI is good at

  • Scores risk + confidence on every finding in seconds
  • Categorises web vs social, urgent vs routine
  • Drafts DMCA notices for known leak-site venues
  • Deduplicates against prior cases automatically
  • Surfaces re-uploads of already-removed pages

Humans do what humans must

  • Confirm the violation, target, and right venue
  • Work platform support relationships on social
  • Sign every notice against authority on file
  • Negotiate with operators, hosts, and registrars
  • Make the call on edge cases the model can't
In the product

Built around the work. No tabs left open.

01 · Triage

AI ranks the damage. Reviewers run the play.

Every finding gets a risk score the moment it lands. Reviewers claim the highest-impact cases first and route social handles to the people who own those platform relationships.

  • Confidence and risk scoring per finding
  • Lane-aware routing: social vs web
  • Reviewer claims, decisions, and audit trail

02 · Evidence

Capture once. Defensible forever.

Screenshots, captured HTML, hashes, and timestamps land in S3 with Object Lock. The page can vanish; the proof can't.

  • Playwright-controlled snapshot per finding
  • SHA-256 content hash + capture metadata
  • Linked to authority record + reviewer decision
Coverage

Two lanes. Both run hard.

Humans run social

Social & messaging takedowns.

Telegram channels, X clones, Reddit reposters, fan-account impersonators. Reviewers own the platform relationships and know which lever moves what — within hours, not weeks.

  • Telegram, Reddit, X, TikTok, IG, YouTube — every finding lands with a reviewer who knows that platform.
  • Reviewers work the platform support channels by hand. No mass-reporting, no scraping, no platform evasion.
  • AI ranks impersonators by risk; the highest-damage handles are claimed first.
AI files, humans verify

Web leak enforcement.

The mechanical part. Discovery, eligibility checks, DMCA notices, follow-up — automated where it's enforceable, with reviewer sign-off on the action that goes out.

  • Known leak-site domains continuously monitored against creator allowlists and authority records.
  • Eligibility checks run automatically: authority on file, live URL, creator match, no allowlist conflict.
  • Notices fan out to operator, host, registrar, or search — whichever removes the page fastest.
AI alone can't make the judgment calls. Humans alone can't move at scale. The whole product is built to combine both — so the notice goes out fast and stands up later.

Founding operator

Barnacle · enforcement lead

Continue the build

Stop chasing leaks.
Start scrubbing them.

Finish the internal workspace, lock creator authority records, and move leak cases through one durable, audit-ready workflow with reviewers handling the channels that need human judgment.

For authorized teams operating on behalf of consenting creators · Social findings remain reviewer-led · We never auto-report or scrape private accounts