Match Group's Tinder scammer reach rose sevenfold, report says

A congressional investigation has put hard numbers on what trust and safety teams long suspected. According to a report published on 13 October, Match Group (MTCH) let Tinder's scam defences go stale while fraudsters adapted. As of late 2025, most of Tinder's machine learning models for spotting bad actors had not been updated in years.

The 37-page report, Outmatched: Tinder, Scammers, and Match Group's Efforts to Catch Up, comes from the Democratic staff of the Joint Economic Committee under Ranking Member Maggie Hassan. It draws on 674 pages of internal documents and an interview with trust and safety chief Yoel Roth. It also credits Match with fixes made since September 2025, when Hassan and Republican Senator Marsha Blackburn opened the probe.

The DII Take

This is a bruising read for Match Group, but not a fatal one. Scammers get through on every platform. The damaging part is that Match's own engineers flagged stale models and rule sprawl in 2025, around the time chief executive Spencer Rascoff told investors that dating had a trust and safety perception issue. Face Check appears to work, but Congress now knows what to ask: how often you retrain, how you measure exposure, and whether your scam data is kept apart from your spam data.

What the investigation found

Internal planning papers from August and September 2025 describe detection propped up by hand-written rules. Staff called many of those rules redundant and out of date, and said bad actors found ways round them quickly. Rising false positives swamped moderators and limited how much Tinder could sample to measure how often real users met bad actors.

Roth told staff that Match retrains its models sometimes yearly, sometimes every couple of years. The experts the committee consulted said that was far too slow, with two of them pointing to monthly retraining as a minimum.

The data Match handed over is harder to argue with:

• The median scammer got fewer than 70 profile impressions before removal in May 2024, and more than 500 by December 2025.
• Almost 1.5 million US spam, bot and scam accounts on Tinder and Hinge matched with a user in 2025 through Q3, roughly a third of all such accounts.
• Nearly 500,000 of them swapped contact details with a user, a step Match itself links to real harm.

Match argued that impressions measure operational efficiency, not user risk. The committee was not persuaded. It said Match named no change to Tinder's recommendation engine to limit how widely suspected scammers are shown before enforcement. Asked for Face Check results on scams alone, Match supplied figures that lumped scams in with spam.

What Match Group has changed

Face Check, a video selfie check for new Tinder users, went nationwide in the US in December 2025. Match told the committee it cut US interactions with bad actors by more than 70% within a month, and that the drop held. Hinge saw a 20% to 30% fall in key markets, and Match.com is next. In the first half of 2026 Tinder added new AI detection models, 169 new automated rules and 108 improved ones.

Hassan said Match had made progress but "it shouldn't have taken so long." Roth told the Boston Globe the report "mischaracterizes both the historical scale of our investment and the work already completed." He also argued that no single company can fix romance scams, and FTC data backs him up. Nearly 60% of people who reported a romance scam loss in 2025 said it began on social media. Reported romance losses hit almost $1.5B last year, against $201M in 2019.

Congress now knows what to ask: how often you retrain, how you measure exposure, and whether your scam data is kept apart from your spam data.

Why every operator should care

The report follows Senate approval of the Romance Scam Prevention Act by unanimous consent on 23 September, according to Fox 17, more than a year after it cleared the House. It applies to any online dating service, whatever its size. When a platform bans an account for likely money fraud, it must warn every user that account messaged, usually within 24 hours.

The FTC will treat breaches as unfair or deceptive practices, and state attorneys general can sue as well. The rules start one year after the bill becomes law and override state notice rules. Smaller operators should build the notification pipeline now, and be ready to explain their own retraining cadence before a regulator asks.

Sources: JEC Minority report: Outmatched (13 October 2026); JEC Democrats press release (13 October 2026); H.R. 2481, Romance Scam Prevention Act (engrossed text); FTC: New data on social media scams (27 April 2026); FTC: Romance scam losses in 2019 (12 February 2020); Boston Globe (14 October 2026); Fox 17 (September 2026)