Investigation: 32 US brokers are selling overlapping datasets with ~2.9B profiles of people pegged as “actively pregnant” or “shopping for maternity products”
Gizmodo identified 32 brokers selling data on 2.9 billion profiles of U.S. residents pegged as … Tweets: @swodinsky , @jjvincent , @swodinsky , @caitlinrcruz , @spekulation , @robinberjon , @andrewcouts , and @swodinsky Tweets: Shoshana Wodinsky / @swodinsky : there were 32 brokers and i spent like *literal days* trying to wrestle *some* sort of intel on data sources from each of them. even then, there were 13 that i just needed to throw up my hands and 🤷🏻♀ ️🤷🏻♀ ️🤷🏻♀ ️🤷🏻♀ ️🤷🏻♀ ️ James Vincent / @jjvincent : the US data broker market is frightening: collecting vast amounts of information; totally unregulated; and able to support all sorts of inferences. particularly frightening are the companies selling data on pregnant users following the demise of Roe v Wade https://gizmodo.com/... Shoshana Wodinsky / @swodinsky : i mean, apps in general aren't the only culprit—they never were. web activity and retail transactions were much more popular targets among the brokers we found (but there's still some...... p significant gaps! this kind of research is *hard*) Caitlin Cruz / @caitlinrcruz : Gizmodo identified 32 brokers selling data on 2.9 billion profiles of U.S. residents pegged as “actively pregnant” or “shopping for maternity products.” https://gizmodo.com/... Spek / @spekulation : We live in hell. https://twitter.com/... Robin Berjon / @robinberjon : “Multiple brokers are likely hawking the same information, as half the world does not live in the United States, and half the world is not pregnant.” It's hard to tell if data brokerage is worse when the data is right or when it's wrong. https://twitter.com/... Andrew Couts / @andrewcouts : well goddamn. @swodinsky @KyleBarr5 and @blakersdozen with the canons and the receipts. https://twitter.com/... Shoshana Wodinsky / @swodinsky : my last major scoop for gizmodo is out 🥲🥲🥲 @KyleBarr5 and i found dozens of brokers selling data from people who were presumed to be pregnant—so we went down the rabbit hole sussing out where this data was coming from turns out period trackers aren't the (only) culprit! https://twitter.com/...
Context & Ripple Effects
The investigation places pregnancy-related targeting inside a broker ecosystem that was already broad enough for Vermont’s registry to list 121 U.S. data brokers. Its finding that brokers infer the label from web activity, retail transactions, and period trackers shows that sensitive classifications do not depend on a single source.
Earlier coverage documented sensitive information flowing from apps through Facebook’s analytics SDK, including signals of an intent to get pregnant. A single broker’s decision to stop selling visits to family-planning centers, such as SafeGraph’s withdrawal, does not address overlapping profile inventories sold elsewhere.
First-order effects
- People labeled as pregnant or maternity shoppers are exposed to resale through multiple broker datasets, rather than a single identifiable seller.
- The 32 brokers face immediate scrutiny over how they source, infer, and market pregnancy-related classifications in a largely unregulated market.
Second-order effects
- App, web, retail, and period-tracker data suppliers become more consequential to broker customers because each can feed the same sensitive audience label.
- Voluntary restrictions by individual sellers have limited reach when overlapping datasets let buyers seek comparable pregnancy-related profiles from other brokers.
Third-order effects
- If inference from ordinary commercial signals remains saleable, privacy risk will increasingly attach to derived health and reproductive categories, not only to explicitly supplied medical data.
- The pattern points toward a data-broker market in which transparency and controls must cover collection, aggregation, and audience inference together rather than individual data sales.
The trend: Sensitive personal-data markets are shifting from selling raw behavioral signals to selling inferred audience categories assembled across many intermediaries.