An identical file
The image on that page is byte-for-byte the file you own. The strongest thing BAN can show, and the least arguable.
SHA-256 · exact matchBrand protection operations
Brand protection that shows its working.
BAN captures a suspicious page as evidence, measures it against the material you own, and puts both in front of you before anything is sent. Every number it reports is one it measured — and it tells you plainly when it measured nothing.
The workflow
One path, end to end. Nothing leaves BAN without a person deciding it should.
-
01
You record what you own
Product photography, logos and page copy, each with a note of where it came from and who says they own it. BAN can read a brand’s own public site to suggest material, and nothing is protected until somebody adopts it.
-
02
A suspicious page arrives
From an operator, or from BAN’s own search of the public web for distinctive runs of your product copy.
-
03
BAN captures it as evidence
A full-page screenshot, the page source, and every image on the page downloaded and hashed — recorded with the time, the final URL after redirects, and the platform signals observed. Stored so it cannot be edited afterwards.
-
04
It is compared, and the comparison is shown
Against your protected material, using the measurements below. The raw figures are always stored and always displayed, so you can disagree with a threshold and still see the number.
-
05
You authorize, or you don’t
The rights holder reviews the evidence and decides. That decision is recorded against the exact evidence it was made on, so what you approved cannot drift afterwards.
-
06
Enforcement, then verified removal
BAN assembles a submission-ready notice and evidence packet, tracks the response, and re-captures the page to establish that the material is actually gone.
How BAN decides something is worth your attention
Three measurements, no model. Each one is arithmetic you can check, and each threshold is a judgement about when something deserves a person’s minute — never a statement that something is infringing.
The same photograph, altered
Resized, re-compressed or lightly cropped. A perceptual hash compares the images structurally and reports how far apart they are, in bits.
64-bit perceptual hash · near match ≤ 10 bits · reviewed to 14Your words on their page
Product copy reproduced elsewhere, measured as the share of your text that appears on their page and the longest run repeated word for word.
shingle containment ≥ 15% · or a verbatim run ≥ 20 wordsWhat BAN does not do
Stated here rather than discovered later. Each of these is a rule the software enforces, not a caveat.
-
It does not search the web by an image.
BAN compares images it has downloaded from a page against your material. It finds pages by their words, then proves image copying on what it finds. Nothing here is reverse image search, and it should not be read as if it were.
-
It does not decide that you have been infringed.
BAN is software and enforcement operations, not a law firm. It reports measurements; a trained operator assesses them; you authorize what gets sent. It draws no legal conclusions of its own.
-
It does not give you a score you cannot check.
Where BAN ranks anything, it stores the named signals that produced the number and shows them added up. There is no model output to take on trust.
-
It does not treat silence as safety.
A search that finds nothing says exactly that, and keeps the captures proving it looked. Finding nothing is a result, not evidence that nothing is out there.
Where BAN is today
BAN is being built in the open and is not yet taking customers. It seems better to tell you where the line is than to let you guess.
Built
Evidence capture, the three comparisons above, rights-holder authorization, notice assembly and removal verification — exercised end to end against real pages. No notice has yet been sent on a live matter.
Built
Separation between customers, enforced in the database rather than in application code, and a structured record of the domains, retailers and sellers you have actually authorized.
Not yet live
Automated discovery — searching the public web for distinctive runs of your product copy — is built and tested, and its search step has not yet run against a live provider. Until it has, BAN will not describe it as monitoring anything.
Get in touch
If your product photography or page copy is being copied and you want to see what the evidence actually looks like, write to hello@ban.ai.