Tinder
On this page · 8 sections
- Where the bill stands, and when the clock starts
- What triggers a notice, and who gets it
- What the notice must say, how and when
- Turning the law into backend work
- How automated scam detection sets your notice volume
- How the FTC is likely to enforce it
- Effort by operator size
- What it means for dating app operators
The Romance Scam Prevention Act (H.R. 2481, with S.841 as the Senate companion) puts a new duty on every US dating service. If you ban or suspend an account because you judge it is likely to try to get money from members by fraud, you must tell every member who received a message from that account. You have 24 hours, with set contents, sent by email or text unless the member has agreed to another channel [1]. The FTC enforces the Act, and state attorneys general can sue as well [1]. The duty starts one year after enactment [1]. You can't do this with a message template alone. You need a link between banned accounts and the people they messaged, clear ban reason codes, and records that show the FTC you met the deadline.
Where the bill stands, and when the clock starts
The House passed H.R. 2481 on 23 June 2025 [3]. The Senate passed the same text without amendment by unanimous consent on 23 September 2026 [2][4]. LegiScan records the bill as presented to the President on 14 October 2026 [2]. The sources don't fully agree on timing. A Pettersen office post dated 6 October said the bill would become law that day unless vetoed [6], but the formal record shows it was presented eight days later [2]. Treat the bill as enrolled and awaiting enactment as of 19 October 2026.
The effective date is "1 year after the date of the enactment" [1], so the deadline will most likely fall in autumn 2027. Count the year from the enactment date, not from Senate passage. Federal law also takes over from state rules here. No state can require, ban or otherwise regulate how dating services notify members about fraud-banned contacts [1]. CBO noted that some states already require similar notices and that most large services already send them nationally [3]. If you only built for one state, you'll need to rebuild to the federal standard.
What triggers a notice, and who gets it
Three definitions set the scope, and each one is narrower or wider than teams tend to assume.
- Fraud ban. This means terminating or suspending an account because, in your judgement, there is "a significant risk" the member will try to get cash or other currency from another member by fraud [1]. Bans for harassment, spam, underage accounts or fake photos alone don't count unless the reason is money fraud. A temporary suspension does count.
- Who must be told. Only members who received a message from the banned member [1]. Members who matched but never got a message, and members who messaged the banned account but got no reply, aren't covered by the text.
- Online dating service. Any website or app that offers members dating or romantic relationships by arranging or facilitating introductions [1]. There is no size threshold, so a niche app with a small member base is covered in the same way as Tinder.
The Act sets no lookback window. In the text, any message received from the banned member triggers a notice, however old it is [1]. That has a direct effect on how long you keep message records.
What the notice must say, how and when
Table 1 maps each statutory requirement to the work it creates.
| Requirement | What the Act says | What to build |
|---|---|---|
| Identify the account | Username or profile identifier of the banned member [1] | Store the display handle at ban time, even if the profile is later wiped |
| Last contact time | Most recent time the recipient sent or received a message to or from the banned member [1] | A per-pair "last message" timestamp, in both directions |
| Fraud warning | Statement, as applicable, that the banned member may have used a false identity or tried to defraud members [1] | Approved template wording, versioned |
| Money warning | Statement not to send cash, other currency or personal financial information [1] | Fixed template block |
| Safety guidance | Best practices to avoid online fraud, which may be a link [1] | A maintained safety page |
| Support contact | Customer service contact details [1] | A monitored inbox or form linked from the notice |
| Channel | Clear and conspicuous, by email or text, or another means if the member consented [1] | A verified email or phone number per member, plus a consent record for push or in-app |
| Deadline | 24 hours from ban initiation, or up to 3 days if the provider judges the circumstances require it [1] | A queue with a service level clock and alerting |
| Law enforcement hold | No notice before a requested delay ends, then send within 3 days [1] | A hold flag with request records and an automatic release |
Source: Romance Scam Prevention Act enrolled text [1].
The channel rule causes the most product work. An in-app inbox message or push notification only qualifies if the member has consented to that channel [1]. Apps that sign members up with social login, or that never verify an email address, need a consent step at onboarding or a verified contact field. The safe harbour helps here. You aren't liable to a member, banned member or former member for a claim based on action taken to comply [1].
Turning the law into backend work
The core job is a graph question: for a given banned account, who did it send messages to, and when was the last exchange? Most teams can answer it from existing data, but not within 24 hours at scale unless they plan for it.
- Conversation edge table. Keep a slim record for each sender and recipient pair: first message, last message in each direction, and the conversation ID. You don't need message content to meet the notice duty. Keeping only this metadata means your retention policy for message content can stay short while the notice duty is still met. Our backend checklist covers where this sits.
- Ban reason codes. Split "fraud ban" from every other ban type in your moderation tooling. Moderators and models must record the reason when they act, because the reason decides whether the 24 hour clock starts.
- Snapshot at ban time. When a fraud ban fires, freeze the recipient list, the banned handle and the per-pair timestamps. Later profile deletion or member clean-up mustn't change what you can prove.
- Notice pipeline. Run a queue that resolves each recipient's contact channel, renders the versioned template, sends it, and records delivery and bounces. Recipients who later deleted their accounts are a grey area in the text, so ask counsel how to treat them.
- Reversals. The Act says nothing about bans that are overturned. If a ban is reversed before notices go out, cancel the pending notices and log why. If notices have already gone out, decide in advance whether to send a correction. Get legal advice on both cases, since a false fraud warning about a real member carries its own risk.
- Compliance log. For each fraud ban, keep: ban ID, initiation timestamp, reason code, decision source (model version and score, or reviewer ID), recipient snapshot, the reason for any delay past 24 hours, any law enforcement request with dates, and the send time, channel, template version and delivery status for each notice.
If you run several brands on one shared member network, check which service each conversation happened on. The duty attaches to "the online dating service" through which the message was received [1].
How automated scam detection sets your notice volume
The law leaves the ban decision to "the judgment of the online dating service provider" [1]. So your detection settings decide how many notices go out. A simple way to plan:
- Notices per period = fraud bans in the period × average distinct members who received a message from each banned account.
Both parts of that sum depend on choices you make. If you catch scammers earlier, at sign-up or before their first message, they will have messaged fewer members by the time you ban them, so you send fewer notices. Pre-message detection therefore cuts notice volume as well as fraud losses. A loose threshold that bans many genuine members creates notices about innocent people and makes reversals more likely. The 3 day window exists for cases where "the circumstances require" more time [1]. It is not a general review buffer, so document the reason every time you use it.
Expect more people to look at these numbers. Reported romance scam losses in the US were $1.14 billion in 2024, with a median loss of $4,000, according to FTC data reported by Dating Industry Insights [5]. For a wider look at prevention, see romance scams on dating apps.
How the FTC is likely to enforce it
A breach counts as a breach of an FTC rule defining an unfair or deceptive act or practice [1]. Because of that, CBO notes that companies that fail to comply "could face civil penalties" [3]. CBO also said it was unclear whether the FTC would seek penalties or other remedies, and that it expected most companies to comply [3]. CBO estimated the FTC would spend $4 million over 2025 to 2030 on guidance, monitoring and enforcement [3]. Expect FTC guidance before the effective date, not a full rulemaking. Then expect early cases that test missed deadlines and incomplete notices.
States can enforce the Act too. A state attorney general must notify the FTC before suing, and the FTC can intervene. No state action can run against a defendant while an FTC case on the same breach is pending [1]. Your timestamped log is your evidence in either forum.
Effort by operator size
These are Dating API's own planning estimates, not sourced figures. The method: we added up the six build items above, assuming an existing moderation tool and email or SMS provider. We then adjusted for the number of brands, the age of the data model and whether ban reasons already exist.
| Operator profile | Main work | Rough effort |
|---|---|---|
| Small, one app, one codebase | Reason codes, edge table, template, simple queue, log | 3 to 6 engineer-weeks, plus legal review of wording |
| Mid-size, several apps or legacy messaging | Backfill of pair timestamps, consent capture, delivery tracking, reversal flow | 8 to 14 engineer-weeks, plus a trust and safety process owner |
| Large, multi-brand or shared network | Cross-brand attribution, high-volume queue, audit tooling, law enforcement hold workflow | 20 engineer-weeks or more across several teams |
Source: Dating API estimate based on the obligations in the enrolled text [1].
If the work exposes deeper gaps, such as messages you can't attribute or ban reasons that were never recorded, read signs your dating app needs a rebuild.
Download the data
Download the tables in this article as CSV files. They are free to reuse with a link to this page.
- Romance Scam Prevention Act: obligations and what to build9 rows · 3 columnsDownload CSV
Questions
Is the Romance Scam Prevention Act law yet?
Both chambers have passed it. The Senate passed H.R. 2481 without amendment on 23 September 2026, and the bill went to the President on 14 October 2026. The duty starts one year after enactment.
What is a fraud ban notification?
It is a notice a dating service must send to a member who received a message from an account it banned or suspended because it judged there was a significant risk of money fraud. The notice has to name the banned profile, give the time of the last message between them, warn against sending money and give a customer service contact.
How quickly must a dating app send the notice?
Within 24 hours of the ban. The provider can extend this to 3 days if it judges the circumstances require it, and law enforcement can ask for a longer delay during an investigation.
Who enforces the Romance Scam Prevention Act?
The FTC, which treats a breach as an unfair or deceptive act or practice. State attorneys general and authorised state consumer protection officers can also bring civil actions. CBO estimated the FTC's costs for guidance, monitoring and enforcement at $4 million over 2025 to 2030.
Does the law apply to small dating apps?
Yes. It applies to anyone in the business of offering a dating website or app that introduces members to each other, and the text sets no size threshold.
