Making the right call here requires understanding the ownership and authenticity of a vast well of products and services, the differences between counterfeiting, reselling, IP infringement, fair use, and reselling, and which name brands or card networks prioritize enforcement on each.

SafetyKit builds general purpose models that understand and action on brand-specific policies, backed by a deep well of counterfeit identification expertise.

Counterfeiting vs. Reselling vs. IP Infringement

The watch on the bottom left is reselling, and it’s both incredibly high counterfeit risk and a massive liability to have on a marketplace. This is a near-perfect photoshop on the Cartier Santos watch, where the ‘Cartier’ logo has been removed and the watch band has been altered.

How many watch experts are also Barbie experts? The phrase “I am Kenough” is trademarked and enforced, making this custom T-Shirt a high risk IP infringement. Had they said “You are Kenough,” the listing would be non-violative, but no font change or tacked-on Ryan Gosling photo would save them. The Nike socks, on the other hand, are tie-dyed: Nike doesn’t make socks in those colors. Had the socks been knock-offs marketed as real Nike, they would be counterfeit, but as the seller is not making any claims like this, the socks are safe to sell.

By now you’ve guessed that the T-Shirt quilt is not any kind of IP infringement, even though it has the sports-brand shirt. The purpose of the shirt has completely changed and it is one small part of a larger quilt.

SafetyKit gets these decisions right.

Detecting counterfeiting, reselling, and IP infringement

The FDA releases new warning letters every week. CITES changes its list of endangered animals. Enforcement trends change quickly: deepfake content and ghost guns are new risks, but a million old policies that could see renewed attention. Paying less attention to non-English listings used to be common practice for merchant monitoring, but Large Language Models have enabled merchant screening in hundreds of languages: “we don’t have a Japanese-speaking review team” isn’t a strong excuse anymore.

SafetyKit sees new risks first, tracks updates to every policy and watchlist, and identifies enforcement trends long before they become policies (and moderates 120 languages out-of-the-box).