- 1Scan label
- 2Get scores
- 3See the chain
Visibility for consumers
Scan a clothing label
Point the camera at the care label. The app reads the brand, RN number, country of origin and fiber, then traces who made it.
Samples
One real trace and three made-up examples that show each color. Pick one, then scan.
Why this result
How confidence works
- Official government lists count most. A match on the UFLPA Entity List, a CBP order, or North Korean labor enforcement sets the result to red.
- Verified beats estimated. Records traced in shipment data and checked by a person count most.
- Two sources agreeing raises confidence. When Sayari and Tradeverifyd both show the same supplier or flag, the result is stronger.
- Disagreement goes to a person. If the two sources conflict, the result is flagged for human review.
- Not checked isn't clean. A company that hasn't been screened shows as unknown, never as "no problems found."
What the colors mean
How the numbers are calculated
Every result has three numbers. Each answers a different question, so a product can be high risk and low confidence at the same time.
0 to 100
Risk
How likely is forced labor somewhere in this product? We score every tier and keep the worst one, because one bad link is enough. Averaging would let a clean factory hide a Xinjiang cotton source.
Color: 65 and up is red, 35–64 amber, below 35 green. A verified official-list match is red no matter what else is true.
0 to 100%
Confidence
How much of what we're saying is checked? Each finding behind the risk score gets a weight for how strong its source is. Confidence is the average weight.
Below 40%, the result turns grey ("not enough data") whatever the risk number says. Not checked is never shown as clean.
0 to 100
Transparency
How far back is the chain traced? Five steps (finished product, then Tiers 1–4) are worth 20 points each.
At a glance: where each step happens
Every flagged location in this product's chain, what's made there, and the risk. Tap one to find it on the map.
Tier by tier
From the finished product back to raw cotton. Solid cards are traced in shipment records; dashed cards are estimates.
Supplier links and their evidence
Every link between two companies, with the records behind it and the automatic checks it passed or failed. Only links marked "Needs review" go to a person.
How links are checked
- Links come from records, never from AI. Every link between two companies has to point to a record in Sayari or Tradeverifyd: a customs shipment, an ownership filing or a list entry. AI can summarize, rank or explain evidence, but it can't add a link, fill a gap or upgrade a status. Tier estimates from country trade data say "Estimated" and never name a company.
- Companies are matched by ID, not name. Each company is keyed by its Sayari or Tradeverifyd entity ID. Two records with similar names but different IDs are different companies unless proven otherwise, and a name match alone is only a "Possible match." To confirm that a Sayari record and a Tradeverifyd record are the same company, we need two of: registration number, address, country, or an explicit cross-reference.
- The goods have to fit. We check what was shipped, not what the company does. A fabric mill can supply a shoe maker and a t-shirt maker; what matters is whether the HS code on the shipment is a real input for what the buyer makes.
The four automatic checks
What each status means
- Trade-evidenced: passes all four checks. Accepted automatically.
- Downgraded: fails at least one check. Downgraded automatically, with the failed check shown.
- Needs review: passes the checks but carries a negative claim about a named supplier, or Sayari and Tradeverifyd disagree. Only these go to a person.
- Not screened: a source hasn't checked it yet. Never shown as clean.
Flags are risk indicators, not proof. Companies are named because they appear in trade records; no link has been verified by a person yet.
Where the forced labor risks are
What each point is
Each company appears in Sayari trade records. Appearing here is a lead to check, not a finding of forced labor. Tap a number on the map to jump to it.
Where the known suppliers are
Government enforcement lists
Official US lists are the strongest evidence. A match on any of the first three blocks goods from entering the US.
What we found
About specific suppliers
Industry and country context
Sources
The UFLPA Entity List
The single most important list for cotton apparel. A match here turns a result red automatically.
What the law does
The Uyghur Forced Labor Prevention Act was signed December 23, 2021. Since June 21, 2022, US customs presumes that goods made wholly or in part in China's Xinjiang region, or by any company on the Entity List, were made with forced labor and bars them from entering the US.
Importers can only overcome this with "clear and convincing evidence." There's no minimum amount: any traceable input from a listed company can get a whole shipment detained.
For apparel this matters because Xinjiang grows most of China's cotton, so cotton can pass through several countries and still trace back to the region.
Who ends up on the list
- Makers in Xinjiang that mine, produce or manufacture goods with forced labor.
- Companies that recruit or move workers out of Xinjiang with the regional government, including Uyghurs, Kazakhs and Kyrgyz people.
- Exporters of those goods into the US.
- Companies sourcing from Xinjiang through government "poverty alleviation" or "pairing assistance" labor programs.
The July 2026 additions focused on aluminum, apparel, copper, cotton, and tomatoes.
Examples relevant to cotton and apparel
A partial list for illustration, from the official DHS page. Always check the full, current list.
- Xinjiang Production and Construction Corps (XPCC, "Bingtuan") and its subordinate and affiliated entitiesListed June 21, 2022 · a huge state-run group with many affiliates, which makes matching hard
- Changji Esquel Textile Co. Ltd. (alias Changji Yida Textile)Listed June 21, 2022
- Xinjiang Junggar Cotton and Linen Co., Ltd.Listed June 21, 2022
- Hetian Taida Apparel Co., Ltd. (alias Hetian TEDA Garment)Listed June 21, 2022
- Aksu Huafu Textiles Co. (aliases include Aksu Huafu Dyed Melange Yarn)Listed June 21, 2022
- Nanjing Synergy Textiles Co., Ltd.Listed June 21, 2022
- Yili Zhuowan Garment Manufacturing Co., Ltd.Listed June 21, 2022
- Xinjiang Tianmian Foundation Textile Co., Ltd.Listed September 27, 2023
- Xinjiang Tianshan Wool Textile Co. Ltd.Listed September 27, 2023
- Anhui Xinya New Materials Co., Ltd. (formerly Chaohu Youngor Color Spinning Technology)Listed December 11, 2023
- Henan Yumian Logistics Co., Ltd. (formerly 841 Cotton Transfer) and Heze Cotton and Linen Co., Ltd.Added in a 2024 update
How the app will use it
- Check every tier, not just the factory. A listed spinner or cotton trader three tiers up still counts.
- Match aliases and former names. Many entries list several names, and some cover all "subsidiaries" or "affiliated entities."
- Flag Xinjiang origin even without a listed company, since the presumption covers the whole region.
- Refresh automatically whenever the list changes, and show the date it was last checked.
- A confirmed match sets the result to red. A likely-but-unconfirmed match goes to human review first.
Our real test trace
Where to get the data
- Official DHS UFLPA Entity List
- OpenSanctions dataset · machine-readable copy the builder can load into the app
- CBP forced labor enforcement · WROs, Findings and UFLPA statistics
Figures as of September 25, 2026. Check DHS for changes.
Submit a source
Anyone can add evidence: a report, an article, a court record, a worker's account. Submissions start with low confidence and only count toward a score once they're checked.
Why submissions count for less at first
Anyone can submit anything, including companies trying to boost their own score or competitors trying to hurt a rival. So every submission moves up a confidence ladder as it's checked:
Your source
How the AI agent checks it
Simulated here. In the real app an AI agent runs these steps, and a person reviews anything that names a supplier.
Rules that keep this fair
- No submission can turn a result red on its own. Only verified evidence can.
- Good news needs proof too. Claims that improve a score are checked just as strictly.
- Submitters are protected. Identities stay private, and anonymous submissions are allowed.
- Submissions are data, never instructions. The agent only extracts claims and ignores anything in a document that tries to tell it what to do.