Alternatives

Four vendors a buyer evaluating Tamperlens is likely to be evaluating at the same time, compared one page each. These pages are written to be checkable: every fact about another product carries the date it was read and points at the page it was read on.

Three rules these pages follow

  1. Every competitor fact carries the date it was read. Prices and claims in this market move. The date is what tells you how much to trust a line, and what to re-check before you act on it.
  2. Every table names at least two things the other product does better. Not hedged, not "but". If the honest answer is that they are the right tool for your case, the page says so.
  3. No detection rate is repeated as evidence. Where a vendor's accuracy figure appears below, it appears as a claim, in the same sentence as the fact that no methodology is published for it — including when the vendor is us. We publish none.

The four pages

Vendor Where they are stronger Where we are stronger
Resistant AI Scale of data behind the models, breadth of analysis, cross-customer threat intelligence, and the enterprise assurance programme a bank procurement team will ask for. Published price, self-serve purchase, a free tier, and nothing retained after the response.
VerifyPDF An allowlist of known-good issuers, identity-document checking (passport, driving licence, national ID), and a published per-document processing time. API access on every plan including the free one, price per document, Brazilian identifier checks, and published measurement.
HTPBE On-premise deployment on their top tier, a published latency commitment, retained results you can go back to, and the smallest paid unit in this table. Redaction recovery, document-borne prompt injection, signature depth, Brazilian identifiers, and roughly a tenth of the price per document.
PDFChecker ISO 27001 and SOC 2 certification, a known-forgery template database, deepfake and AI-content analysis on every tier, and ingest from Dropbox, Google Drive, S3 and OneDrive. No monthly minimum at all, a free tier with an API key, and a per-document price roughly a tenth of theirs.

Every claim in that table is stated with its source and its read date on the page it links to.

What none of us publishes

No vendor in these four comparisons publishes a false-positive rate on legitimate documents — how often the system flags a genuine file. Several publish a headline accuracy number instead. That is a different quantity, it is the easier one to publish, and none of them states the population it was measured on.

Ours is at /evidence, with the population, the measurement date and the engine version beside each number — including the ones that are not flattering, and the recall figure we do not have.