The Law Guarantees Your Cancel Button. It Does Not Let You Enforce It. Your Bank Does.

You Cancelled. The Charge Came Anyway.

Four different powers can stop a recurring charge. Only one of them belongs to you.

You clicked through the maze. You found the page that was three menus deep. You answered the “are you sure?” three times, declined the discount, and got a confirmation screen. Then the next month arrived, and so did the charge.

You may have read that this is illegal now — that a federal rule makes cancelling as easy as signing up. That rule is real, and it is dead. A court erased it in July 2025, six days before it was supposed to start. We told that story in our guide to disputing a credit card charge, and this article takes it as settled. This one asks the question that comes next.[1]

If the famous rule is gone, what is actually left? And of what is left, how much of it can you personally use? Those turn out to be two very different questions, and almost nobody answers the second one.

The short answer is the shape of this whole article. Four separate powers can stop a recurring charge in the United States. The federal rule belongs to nobody — it reverted to a mail-order regulation written in 1973. The federal statute is alive and has teeth, but it belongs to the government, not to you. A state law might belong to you, depending entirely on your address. And your bank holds a switch that works in every state, without the seller’s permission, and that almost no one uses.

One more thing before we start, because it changes how urgent this feels. A forgotten subscription is not a small leak. It is a payment you make on purpose, forever, without deciding to. Fifteen dollars a month is a hundred and eighty dollars a year, and it renews on its own for as long as the card keeps working.

A Rule Is Not a Law, and That Difference Decides Everything

Agencies write rules. Congress writes laws. A court can erase a rule in an afternoon.

Almost every confusing headline about subscriptions comes from one blurred word. People say “the law” when they mean two completely different things, and those two things have completely different life expectancies.

A statute is written by Congress and signed by a president. To change one, you need another act of Congress. A rule — also called a regulation — is written by an agency like the Federal Trade Commission, under power that a statute handed it. Rules can do things statutes cannot: they get specific, they name deadlines, they describe buttons. But they are borrowed power, and borrowed power can be taken back. A federal court can vacate a rule and it simply stops existing.[2, 3]

Hold that distinction and the last two years stop being confusing. The thing that made headlines and then died was a rule. The thing that took two and a half billion dollars out of Amazon is a statute. They were never the same object. One of them was always going to be fragile, and one of them never moved at all.

But there is a second question hiding behind the first, and it is the one this article is really about. Even a living statute is only useful to you if you are allowed to pick it up. Some laws hand you a weapon. Some laws hand a weapon to a government lawyer and give you a phone number. That difference is invisible in every headline, and it is the difference between “I have rights” and “I can do something today.”

So here is the map. Four powers, and the only honest way to read it is by asking who is holding each one. The federal rule: nobody holds it, because it is a mail-order regulation from 1973. The federal statute: the FTC holds it, and so do state attorneys general — you do not. Your state’s auto-renewal law: you might hold it, if you live in the right place, and even then it depends which state. Your bank: you hold it, everywhere, and it is the only one on this list that is unambiguously yours.

Open the Rulebook Today and You Find a Rule About Mailing People Books

The Negative Option Rule still exists. It was written in 1973, and it has nothing to do with your streaming service.

Here is something almost no article does: actually open the rulebook and read what is printed there this morning. The Federal Trade Commission’s subscription rule lives at Part 425 of Title 16 of the Code of Federal Regulations. Its title today is “Use of Prenotification Negative Option Plans.”[4]

A “prenotification plan” is the book-of-the-month club. The seller mails you an announcement saying which book is coming. If you do not mail back a form in time, the book ships and you are billed. The rule is a machine for regulating that, and only that. It talks about postage. It talks about the Postal Service. It gives you “at least ten days in which to mail your form.” It cares whether your form was postmarked three days before the return date.[5]

Now search that rule for the words that describe your actual life. “Automatic renewal” appears zero times. “Free trial” appears zero times. “Simple mechanism” — zero. “Express informed consent” — zero. “Recurring subscription” — zero. Section 425.2, where a modern rule might have gone, reads in its entirety: [Reserved].[4, 5, 6]

You do not have to take our word for how narrow this is, because the FTC says it out loud. In its own 2026 notice, the Commission writes that the Rule “applies only to prenotification plans for the sale of goods and does not reach most modern negative option marketing.” That is the agency describing its own rulebook as beside the point.[7, 8]

And then there is the detail that tells the whole story in six characters. Every part of the CFR carries a source note saying when its text was published. Part 425’s reads: “91 FR 6509, Feb. 12, 2026.” February 2026 is recent. It is one of the newest datelines in this corner of the rulebook. The words it stamps are from 1973. The freshest text in the file is the oldest rule in it.[4, 9]

How the 1973 text got a 2026 dateline is the next section, and it is not the story most people think it is.

The Rule Died Over a Missing Piece of Paperwork, Not Over You

The court never decided whether an easy cancel button is a good idea. It never got that far.

In November 2024 the FTC published the rule everyone remembers: disclose the terms, get real consent, and give people a way out that is as easy as the way in. Industry groups asked four different federal appeals courts to kill it. A panel that handles multi-court pileups consolidated all the petitions into the Eighth Circuit — which is to say, the venue was close to a coin flip.[10, 9]

On July 8, 2025, the Eighth Circuit vacated the rule. Read the opinion and you find something the headlines skipped entirely. The court did not rule that the cancel-button requirement was wrong, or unfair, or beyond the agency’s power. It wrote that the Commission “failed to follow procedural requirements under § 22 of the Federal Trade Commission Act” — and vacated on that basis alone.[1, 11]

The missing paperwork was something called a preliminary regulatory analysis. Federal law requires one when a proposed FTC rule would have an effect of $100 million or more on the national economy. The FTC skipped it, because it had “preliminarily determined” the rule would not reach $100 million. Compliance, it reasoned, “should not create any substantial added burden,” since most sellers already disclosed something and let people cancel somehow.[1, 11]

Here is the part worth sitting with. The FTC’s own administrative law judge disagreed with the FTC. As the appeals court recorded it, the ALJ “found the Rule would meet the $100 million economic impact threshold, even though the Commission initially estimated it would not.” The agency’s own hearing officer said the rule mattered more than the agency had claimed — and that finding is what made the missing analysis fatal.[1]

The FTC argued the slip was harmless. The court rejected that too. And so the most consequential consumer rule of the decade was undone not because anyone decided it was bad policy, but because the agency underestimated its own rule’s importance and skipped a step. It never applied to anyone for a single day.[1]

Seven months later the FTC made it official in the rulebook. On February 12, 2026, it published a final rule to “recodify the text of the Negative Option Rule as it existed before the effective date of the Commission’s 2024 final rule.” Effective immediately. That is the sentence that put 1973 back on the page — and put a 2026 dateline on top of it.[9, 12]

The Law That Survived, and the One-Line Reason It Could Not Be Touched

ROSCA borrows its key definition from a different rulebook. That accident of drafting is why it is still standing.

While the rule was being written and unwritten, a statute sat quietly in the U.S. Code doing exactly what the rule was supposed to do. Congress passed it in December 2010 and gave it a name nobody remembers: the Restore Online Shoppers’ Confidence Act. Everyone calls it ROSCA.[13, 14]

ROSCA makes it unlawful to charge anyone for goods or services sold online through a negative option feature unless the seller does three things. Disclose all material terms before taking your billing information. Get your express informed consent before charging. And provide “simple mechanisms for a consumer to stop recurring charges.” That last phrase is the cancel button, written into a statute, fifteen years ago.[14]

So why did the Eighth Circuit’s wrecking ball miss it entirely? Because of one parenthesis. ROSCA does not define “negative option feature” itself. It borrows the definition — and it borrows it from the Telemarketing Sales Rule at 16 CFR Part 310, not from Part 425. The statute literally points at a different rulebook.[14, 15]

Go look at what it points to. Section 310.2 defines a negative option feature as a provision “in an offer or agreement to sell or provide any goods or services” under which “the customer’s silence or failure to take an affirmative action to reject goods or services or to cancel the agreement is interpreted by the seller as acceptance.” That definition is alive. It was never challenged. And it says services.[15, 16, 17]

That is the whole trick. Part 425 covers merchandise sent by mail and nothing else. ROSCA covers goods or services sold on the internet — streaming, software, gyms billed through an app, anything. Vacating Part 425 did not narrow ROSCA by a single word, because ROSCA was never leaning on Part 425 in the first place.[5, 14]

ROSCA has a second half worth knowing about, too. Section 8402 goes after the “post-transaction third party seller” — the outfit that pops up after you buy something from someone else and enrolls you in its own program using billing details it got handed. That practice is why the statute exists at all.[18]

Read One More Sentence and the Law Stops Being Yours

ROSCA says you get a simple way to cancel. It does not say you can sue about it.

Most coverage stops at “ROSCA requires simple mechanisms,” and it feels like an ending. You have a right, the right is in a statute, the statute is alive. Read one section further and the feeling changes.[14]

Section 8404 is titled “Enforcement by Federal Trade Commission.” It says a ROSCA violation “shall be treated as a violation of a rule” under the FTC Act, and then: “The Federal Trade Commission shall enforce this chapter.” Section 8405 is titled “Enforcement by State attorneys general.” It lets a state AG sue in federal court, on behalf of that state’s residents, for an injunction — after giving the FTC written notice, and with the FTC free to step in.[19, 20]

Those are the only two enforcement doors ROSCA builds. The FTC, and a state attorney general. Search the chapter for a third and there is not one. There is no section that lets a consumer file a case, no damages formula, no attorney-fee provision, nothing. Lawyers call this having no private right of action.[19, 20]

This shape shows up all over American consumer law, and it catches people every time. Airline passengers hit the same wall: the statute that bans unfair airline practices has no passenger lawsuit in it either, as we explain in our guide to airline refunds and delay compensation. Workers hit it with OSHA retaliation. The pattern is always the same — Congress writes a duty, and hands the key to a government lawyer.

So both federal doors are closed to you personally. The rule is a 1973 relic. The statute is real but belongs to someone else. If you stop reading here, the honest summary is bleak: on paper you are protected, and in practice you cannot personally do a thing with it.

Do not stop here. The two doors that are left are the ones this article was written for, and one of them is open to everyone. But first, the closed doors are worth using for what they can do — because they are not decorative.

You Cannot Sue Under It. You Can Still Feed It.

The complaint you file is not a case. It is ammunition for the only people who can bring one.

A law you cannot personally invoke is not a law that does nothing. It is a law that runs on volume. The FTC does not go hunting for one person’s gym membership; it moves when a pattern becomes undeniable, and patterns are built out of complaints.

When it does move, it is not gentle. In June 2023 the FTC sued Amazon under ROSCA over Prime sign-ups and cancellations. In the agency’s own summary of the case, the charge is blunt: “The primary purpose of its Prime cancellation process was not to enable subscribers to cancel, but to stop them.” On September 25, 2025, the court entered a stipulated order. Two and a half billion dollars.[21, 22, 23]

Notice the date. The rule everybody was mourning had been dead for seventy-nine days when that order landed. Nothing about the vacatur slowed it down, because the case was never built on the rule. It was built on the statute, and the statute never moved.[22, 1]

The settlement also proves the “you get nothing” instinct wrong. The order sent money back to consumers, and the FTC runs a page telling people how to get it. That is what a law you cannot personally sue under looks like when it works: someone else does the suing, and a refund shows up in your account.[24, 25]

So your move under ROSCA is a report, and it takes five minutes. File it at reportfraud.ftc.gov. Then file the same facts with your state attorney general, because section 8405 makes that office one of exactly two that can walk into federal court holding this statute. Say what you were charged, when you tried to cancel, and what the screen did. You are not filing a case. You are handing evidence to the people who can.[26, 20]

That is worth doing. It is not worth waiting for. The next three sections are about the levers that move this month.

Click to Cancel Did Not Die. It Moved to the States.

While Washington was resetting to zero, two of the biggest states wrote the dead rule into their own law.

Line up the dates and a story appears that almost nobody has told. On July 1, 2025, California’s overhauled auto-renewal rules became operative. Seven days later, on July 8, 2025, the Eighth Circuit vacated the federal rule. Four months after that, in November 2025, New York’s rewritten auto-renewal statute took effect — containing, almost word for word, the requirement the federal rule had died carrying.[27, 28, 1]

New York’s section 527-a is the clearest example, so read what it actually demands. A business may not “fail to provide the consumer with the option to cancel at any time using a simple cancellation mechanism that is as easy to use as the mechanism that the consumer used to provide consent and that is through the same medium that the consumer used to provide consent.” If you subscribed with two clicks on a website, two clicks on that website is the standard.[28]

It goes further than the federal rule ever did. The statute enumerates dark patterns by name and forbids them: a business may not obstruct or unreasonably delay a cancellation, and “unreasonable or unlawful conditions include, but without limitation, hanging up on consumers who call to cancel, providing false information about how to cancel, misrepresenting the consequences or costs of cancellation, or misrepresenting the reasons for delays.” That is a retention script, described in a statute, and banned.[28]

It even settles the argument about the “wait, here is 50% off” screen. New York allows it — a business “may present the consumer with a discounted offer, retention benefit or information regarding the effect of cancellation” — but it may not use that moment to obstruct you. The save offer is legal. The maze around it is not.[28]

None of that language was in New York’s law six months earlier. Pull the older version of the same section and it talks about a “toll-free telephone number” and letting people who signed up online cancel “exclusively online” — the weaker, older generation of these statutes. The as-easy-as-signup standard, the dark-pattern list, and the retention rule all arrived in the 2025 rewrite. New York did not react to the vacatur so much as walk straight through the hole it left.[28]

Here is the honest catch, and it matters more than the good news. These laws are not interchangeable, and most of the country is not New York or California. There is no reliable public count of which states have a real auto-renewal law, how strong it is, or who can enforce it — so do not trust any article that gives you a tidy number, including a number about how many states are covered. Search your own state’s attorney general for “automatic renewal,” and find out what you actually have before you count on it.

California Wrote an Anti-Dark-Pattern Clause Into a Statute

And unlike New York, it lets you do something about it yourself.

California’s auto-renewal law is the one most companies actually design around, because California is too big to build a separate checkout for. Its 2024 overhaul, AB 2863, is codified in the Business and Professions Code, and the section itself records its own pedigree: “Stats. 2024, Ch. 515.” If you see “Chapter 936” for this bill online, that is wrong; the codified section is the thing to read.[27, 29]

The provisions read like a checklist of the tricks. Terms must be clear and conspicuous before the agreement is fulfilled and in visual proximity to the request for consent. A free trial must explain, up front, the price that hits when it ends. The business must get affirmative consent before charging, and must send an acknowledgment that includes the cancellation policy in a form you can keep.[27]

Then comes the clause that is genuinely unusual. It is unlawful to “include any information in the contract that interferes with, detracts from, contradicts, or otherwise undermines the ability of consumers to provide their affirmative consent.” That is a dark pattern described in the abstract and outlawed in the abstract — no list of specific tricks to route around. California also makes the seller keep proof of your consent for at least three years, which quietly answers the “our records show you agreed” conversation.[27]

And here is where California and New York part company, which is the part that decides whether the law is yours. New York’s section is enforced by the attorney general: the statute lets the AG apply for an injunction and get restitution, and that is the mechanism it names. California instead says that a violation is not a crime, but “all available civil remedies that apply to a violation of this article may be employed” — with a safe harbor for a business that complies in good faith. Same idea, very different question of who is holding it.[30, 28]

Both states also carve out whole industries, which is the kind of detail that decides whether the law helps you at all. California exempts businesses regulated by the state utilities commission, the FCC or FERC, among others. New York exempts entities regulated by its financial services department, banks and credit unions, alarm companies, and service-contract sellers. So the auto-renewal law that would rescue you from a streaming service may say nothing at all about your alarm monitoring or your phone line.[31, 28]

That is the practical lesson of this whole act, and it is why “check your state” is not filler advice. Two of the strongest auto-renewal laws in the country agree on what a business must do and disagree on whether you personally can do anything when it does not. Your zip code is doing more work here than the statute is.

Fifty-Five Years of Consumer Law Have Been Protecting Boxes

Every law that hands you a free-gift remedy stops at the doorstep. The thing charging you every month never arrives at the door.

There is a pattern buried under everything above, and once you see it you cannot unsee it. American consumer law has an unusually generous remedy for being charged for something you never agreed to: you keep it, for free, and you do not even have to mail it back. That remedy is more than half a century old, and every single version of it has the same blind spot.

Start in 1970. Federal law says that merchandise mailed to you without your prior express request “may be treated as a gift by the recipient, who shall have the right to retain, use, discard, or dispose of it in any manner he sees fit without any obligation whatsoever to the sender.” The FTC still tells consumers this in plain language today: “You’re legally entitled to keep it as a free gift.”[32, 33]

Then 1973: the Negative Option Rule, which we have already read. Goods, by mail. Then 2009: California says that if a business sends “goods, wares, merchandise, or products” under an auto-renewal without first getting your affirmative consent, they are “deemed an unconditional gift,” shipping costs included. Then 2020: New York enacts the identical remedy, using the identical four words.[5, 34, 28, 35]

Now say the list out loud. 1970: “merchandise mailed.” 1973: goods, by mail. 2009: goods, wares, merchandise, products. 2020: goods, wares, merchandise, products. Fifty years of lawmaking, four different legislatures, and the strongest remedy any of them offers only fires when something physical shows up at your door.

The strangest part is that at least one of those legislatures knew better. California opens its auto-renewal article by stating its own purpose: to end the practice of charging cards “without the consumers’ explicit consent for ongoing shipments of a product or ongoing deliveries of service.” Services are named in the very first line. Four sections later, when the statute reaches for its sharpest tool, the tool only grips “goods, wares, merchandise, or products.” The intent covers your streaming subscription. The remedy does not.[36, 34]

Your streaming service never shows up at your door. Neither does the gym, the cloud storage, the app, the newsletter, or the software seat. The single federal law on this list that says the word services is ROSCA — the 2010 statute — and the definition it borrows from the Telemarketing Sales Rule reaches “any goods or services.” That is the whole reason ROSCA matters, and it is the reason it is the only federal thing standing.[14, 15]

So the country wrote its best consumer remedies for a world of book clubs and mail-order steak, and then the economy quietly stopped shipping anything. That is not a conspiracy. It is just old law. But it explains why the honest answer to “what protects me?” keeps landing somewhere other than the statute books — and why the next section is about your bank.

Your Bank Has a Switch, and the Seller Does Not Get a Vote

Three business days before the charge, orally or in writing. It works in every state, and it does not care what the cancel page says.

Every door so far has depended on somebody else deciding to act — a court, an agency, a state legislature, the company itself. This one does not. If the money leaves your bank account by electronic transfer, federal banking rules give you a right that runs against your own bank, and your bank has to do it.

The rule is Regulation E, section 1005.10, and the operative sentence is short. “A consumer may stop payment of a preauthorized electronic fund transfer from the consumer’s account by notifying the financial institution orally or in writing at least three business days before the scheduled date of the transfer.” Read it slowly, because four separate things in it matter.[37, 38, 39]

“A consumer may.” Not the FTC. Not an attorney general. You. This is the first thing in this entire article that is written in your name. “Orally or in writing.” A phone call is legally sufficient to start it — you do not need a lawyer, a letter, or a form. “At least three business days before.” That is the whole deadline. “Preauthorized electronic fund transfer” means a transfer out of your account set to recur at substantially regular intervals — which is exactly what a subscription is.[38, 40]

Notice what is not in that sentence. It does not say the seller has to agree. It does not say you must have cancelled first, or won an argument, or proved anything. It does not ask whether the charge was fair. Regulation E simply lets you tell your own bank to stop moving your money, and the seller is not part of the conversation.[38]

This is also the answer to a question our guide to overdraft fees raises and deliberately leaves open. The overdraft opt-in shield covers debit card and ATM transactions — it does not cover recurring ACH debits like a gym membership or a utility autopay, so a bank can pay those into the negative and charge you for it whether you opted in or not. Regulation E’s opt-in rule cannot help you there. Regulation E’s stop-payment rule can.[37]

Two practical notes before the trap in the next section. Banks generally charge a fee for a stop-payment order — it is one of the fees that survives on “free checking,” which we cover in how to choose a bank account. And the CFPB’s own consumer page on stopping automatic payments tells you to call and then follow up in writing, which is good advice that stops short of the regulation’s actual deadlines. The deadlines are the part that decides whether it works.[41]

The Fourteen-Day Trap That Brings the Charge Back

You call, the bank stops it, and two weeks later it is charging again. This is written into the rule, and the bank has to warn you.

The phone call is the easy part, and it is also where this quietly falls apart for people. Regulation E lets your bank ask for more than the call. Subsection (c)(2): the institution “may require the consumer to give written confirmation of a stop-payment order within 14 days of an oral notification.”[38]

And if you do not send it, the rule says exactly what happens, in one flat sentence: “An oral stop-payment order ceases to be binding after 14 days if the consumer fails to provide the required written confirmation.” Your stop dies on day fifteen. Nobody calls to tell you. The next charge simply goes through, and it looks like the bank ignored you.[38, 37]

There is a protection built in, and it is worth knowing because it turns this from a trap into a checklist. If the bank is going to require written confirmation, the rule says it “shall inform the consumer of the requirement and provide the address where confirmation must be sent when the consumer gives the oral notification.” Not later. On that same call.[38]

So the script is two sentences long. Call and say you are stopping payment on a preauthorized transfer. Then ask: “Do you require written confirmation, and what address does it go to?” If they say yes, you now have fourteen days and a mailing address, and you send it the same day. If they say no, note the date and the name of the person who said it.

Two neighbouring rules are worth pocketing while you are here. If the amount of a recurring transfer is going to change from last time, the payee or the bank must send you written notice of the new amount and date at least ten days before it happens — which is why a silent price increase on an autopay is not just annoying, it is a rule problem. And a preauthorized transfer is supposed to be set up on a written authorization signed by you, with a copy given to you. If a gym has never handed you one, that is worth raising.[38, 42]

Stopping the Payment Is Not the Same as Ending the Contract

The switch protects your account. It does not erase what you agreed to — and the gap between those two is where the collection letter lives.

This is the sentence that keeps the last section from turning into bad advice. A stop-payment order is a banking instruction. It tells your bank not to move money. It does not tell the seller anything, and it does not touch the agreement you signed.

So if you owe the money — a real contract, a term you agreed to, a service you are still using — stopping the payment does not make the debt disappear. It makes the debt unpaid. That is a different problem, and it can travel: late fees, a suspended account, and eventually a collection agency. If it gets that far, your rights change hands entirely, and our guide to debt collectors and the FDCPA is the one you want.

The stop-payment switch is at its strongest when the money is the thing that is wrong: you already cancelled and they charged anyway, the trial converted at a price you were never shown, the “membership” renewed after you moved cities. In those cases you are not dodging a bill. You are stopping a transfer you never validly authorized in the first place, and you are doing it while the paperwork catches up.

Which means the right order is: cancel first and keep the proof, then stop the payment as a backstop. Do not use the switch instead of cancelling. Use it because cancelling is a request and a stop-payment order is an instruction, and only one of those is addressed to someone who works for you.

If It Is on a Credit Card, You Have a Third Door

Different card, different law, different article.

Everything in the last three sections assumed the money leaves a bank account. If the subscription rides on a credit card instead, you are in a different rulebook with a different clock, and it is genuinely one of the better consumer laws on the books.

We are not going to re-teach it here, because we already wrote it properly. Our guide to disputing a credit card charge covers the deadline that actually matters, the letter that almost nobody sends, why a new card number does not save you, and the reason the funding source you pick decides which rules you get. Read that one and come back.

One boundary is worth drawing clearly, because these three get mixed up constantly and they are three different sections of law. A charge you never authorized — someone else got into your account — is an error-resolution problem, and our guide to Zelle, Venmo and Cash App scams owns that ground. A charge you did authorize but want to end is this article: Regulation E, section 1005.10(c). A charge on a credit card you want reversed is the dispute guide. Same wallet, three different rulebooks.[39, 43]

The Two Moments They Get You: the Trial Ending and the Exit

One is a design decision about the day you stop paying attention. The other is a design decision about the day you try to leave.

A free trial is not a gift. It is a scheduled charge with a delay on it, and the whole design question is whether you will remember. The state laws we just read attack exactly this: California requires the offer to explain, before you agree, the price that will hit when the trial ends, and New York requires the same clear-and-conspicuous explanation of how and when the price changes.[27, 28]

California even has a name for the moment: a “free-to-pay conversion,” written into the statutory definition of an automatic renewal. Naming a thing in a definitions section is how a legislature says it has seen this before. And the same law defines what “clear and conspicuous” has to look like so it cannot be argued away — New York spells it out as larger type, or contrasting type, font or colour, or set off by marks, “in a manner that clearly calls attention to the language.”[44, 45]

Notice what that tells you about the norm. Legislatures do not write rules banning things nobody does. The reason two of the largest states independently decided to require a plain sentence about the post-trial price is that the plain sentence was routinely missing.

The second moment is the exit, and the FTC has already described it under oath. Its case against Amazon accuses the company of using “manipulative, coercive, or deceptive user-interface designs known as dark patterns” to enroll people, and says leadership “slowed or rejected changes that would’ve made it easier for users to cancel Prime because those changes adversely affected Amazon’s bottom line.” That is not a metaphor about bad UX. That is an allegation that the friction was the product.[21, 46]

Once you have read New York’s list, you can name the moves as they happen. Hanging up on a caller who says the word “cancel.” Giving false information about how to cancel. Misrepresenting what cancelling will cost you or what you will lose. Inventing a reason the request is taking so long. New York wrote those four down because they are a script, and scripts repeat.[28]

The practical upshot is unglamorous: assume the exit is designed, and document it while you are inside it. A screenshot of the maze is worth more later than your memory of it, and in a state with a dark-pattern statute it is worth more than that.

Before You Stop Anything, Find Out What Is Actually Running

The charges you remember are not the problem. The problem is billed through a name you do not recognise.

Everything above is useless against a subscription you have forgotten exists, and forgotten subscriptions are the expensive ones. They do not announce themselves. They arrive as a nine-character merchant string on line forty of a statement you scroll past.

So do the boring thing first: pull twelve months of statements, not one. A monthly scan misses everything annual, and annual renewals are where the biggest single charges hide — the one you agreed to on a discount fourteen months ago and have not thought about since.

Then check three places, not one, because a subscription can be billed through any of them and each hides from the others. Your bank account for ACH debits. Your credit cards, each one separately, including the one you keep for travel. And your app store accounts — Apple and Google both keep their own subscription lists, and a charge billed through a phone platform will not appear as the service’s name anywhere on your card statement.

That last one matters for a reason beyond bookkeeping: where a subscription is billed decides which door you get. Billed to your checking account, and the stop-payment right in this article is available. Billed to a credit card, and the dispute guide applies. Billed through an app store, and the platform’s own cancellation flow is usually the only thing that actually ends it — the service itself often cannot.

Do the sum at the end and write the number down. Not the monthly number — the annual one. Twelve dollars a month is not a decision anyone would defend if it were presented once a year as a hundred and forty-four dollar bill with a renewal box already ticked.

How to Cancel So That It Sticks

Six steps, in this order. The order is the whole point.

One: cancel through the same channel you joined through, and do it in writing if you can. If you signed up on a website, cancel on that website. That is not just tidy — in New York it is the legal standard, and in California the acknowledgment you were sent is supposed to tell you how. A chat transcript or an email is evidence. A phone call, by itself, is a memory.[28, 27]

Two: screenshot the whole path, not just the end. The confirmation number matters, but so does the maze — the retention offer, the phone number they sent you to, the page that timed out. In a state with a dark-pattern statute, that trail is the case. Capture the date on screen.

Three: find the next scheduled charge date, and count backwards three business days. This is the step everybody skips, and it is the one with a deadline in it. Regulation E gives you a stop-payment right only if you notify the bank at least three business days ahead. Business days, not days — a charge on Monday means a call by Wednesday of the week before, and a holiday moves it.[38]

Four: call the bank and ask the one question. Say you want to stop payment on a preauthorized electronic fund transfer, name the merchant and the amount, and then ask whether they require written confirmation and where it goes. If they do, send it that day — you have fourteen, and on day fifteen an unconfirmed oral order stops binding anyone.[38]

Five: watch the next two cycles, not the next one. A stop-payment order that lapsed on day fifteen looks exactly like a stop-payment order that worked — right up until the month after. Check the account on the date the charge would have landed, twice.

Six: if it charges again, do not re-argue with the merchant. Escalate. You already did the cancelling. The next section is the ladder.

When They Keep Charging Anyway: the Ladder

Five rungs, cheapest first. Most people stop at the bottom one and give up.

Rung one — the merchant, once, in writing. One message that says the date you cancelled, the confirmation you got, and the charge that arrived anyway. Not a negotiation. A record. If they fix it, you are done and you have a paper trail; if they do not, rung one just became evidence for every rung above it.

Rung two — the bank, on the clock. Stop payment, three business days ahead, and answer the written-confirmation question. This is the rung that actually stops the money, and it is the only one that does not depend on anyone else agreeing with you.[38]

Rung three — the card issuer, if it is a credit card. Different law, different deadline, and it has its own guide. Do not let the clock on that one run out while you are still emailing the merchant.

Rung four — the regulators, both of them. The FTC at reportfraud.ftc.gov, and your state attorney general. Neither will call you back about your gym. Both are building the pattern that produced a two-and-a-half-billion-dollar order against the largest retailer in the country, and section 8405 makes your AG one of exactly two offices that can walk this statute into federal court.[26, 20, 22]

Rung five — small claims court. You cannot sue under ROSCA. You can sue on your contract, and in California you may have state civil remedies on top. Small claims is cheap, has no lawyers, and a subscription case is unusually clean: here is what I paid, here is when I cancelled, here is the screenshot. Our guide to small claims court walks the filing.[30]

What Changes Next, and Why You Should Not Wait for It

The FTC restarted in March 2026. It is at the stage before the stage before a rule.

On March 13, 2026, the FTC opened a new proceeding on the Negative Option Rule. The summary says exactly the right thing: the Commission wants comment on amendments to “help consumers avoid recurring payments for products and services they did not intend to order and to allow them to cancel such payments without unwarranted obstacles.” That is the whole problem, stated by the agency, in 2026.[7, 8]

Now read the label on the document, because it is the part that matters. It is an Advance Notice of Proposed Rulemaking. Not a rule. Not even a proposed rule. It is the notice that comes before the proposal — the step where an agency asks whether it should start drafting at all. Comments closed on April 13, 2026.[7]

As of the date on this article, no proposed rule has followed. And the road from here is the same road that just failed: a proposal, a comment period, a regulatory analysis this time, a final rule, an effective date, and then whatever the courts do with it. The last attempt took from 2023 to 2024 to write and died in 2025 without ever applying to anyone.[47, 48, 1]

Pressure is coming from both directions, which is worth knowing if you are trying to guess. Consumer groups petitioned the agency in December 2025. A separate petition went the other way in January 2025. The Commission has been asked to move and asked to stop, and it has answered by starting over at the earliest possible stage.[49, 50]

So the honest forecast is: not soon, and not certainly. Which is the argument for the whole shape of this article. The rule may come back in some form, someday. Your bank’s stop-payment right is in the rulebook this morning, it has been there for decades, and it does not need anyone’s permission — including the seller’s.[37]

Key Takeaways

The famous rule is gone, and what replaced it is a 1973 mail-order rule. Part 425 today governs prenotification plans — the book-of-the-month club. Its own agency says it “does not reach most modern negative option marketing.” Searching it for “automatic renewal” or “free trial” returns nothing.[4, 7]

It died on paperwork, not on principle. The Eighth Circuit vacated it because the FTC skipped a preliminary regulatory analysis — after the agency’s own ALJ found the rule would cross the $100 million threshold the FTC said it would not. No court has ruled against easy cancellation.[1]

ROSCA is alive because it borrowed its definition from a different rulebook. The statute points at the Telemarketing Sales Rule, 16 CFR 310.2(w), which reaches “any goods or services.” That is why vacating Part 425 changed nothing about it — and why the FTC could take $2.5 billion from Amazon seventy-nine days after the rule died.[14, 15, 22]

ROSCA is not yours. Section 8404 gives enforcement to the FTC; section 8405 gives it to state attorneys general. There is no private right of action anywhere in the chapter. Your move under ROSCA is a report to reportfraud.ftc.gov and to your state AG — real, worth doing, and not a substitute for acting yourself.[19, 20, 26]

Click to cancel moved to the states. New York’s rewritten section 527-a now requires a cancellation mechanism “as easy to use as the mechanism that the consumer used to provide consent” and “through the same medium,” and it bans hanging up on people who call to cancel. California’s AB 2863 rules went operative one week before the federal rule died. But New York’s is enforced by the attorney general, while California allows “all available civil remedies” — so your zip code decides whether the law is yours.[28, 27, 30]

The strongest remedies were all written for boxes. The 1970 unordered-merchandise law, the 1973 rule, California’s unconditional-gift clause and New York’s all stop at physical goods. Only ROSCA says “services.” If it never arrives at your door, most of American consumer law never sees it.[32, 34, 14]

Your bank is the only lever that is unambiguously yours. Regulation E, section 1005.10(c): stop payment on a preauthorized transfer by telling your bank, orally or in writing, at least three business days before the charge. No state law required. No seller cooperation required.[38, 39]

Ask one question on that call, or the stop dies. The bank may require written confirmation within 14 days, and without it “an oral stop-payment order ceases to be binding after 14 days.” The bank must tell you and give you the address when you call. Ask, then send it the same day.[38]

Stopping the payment is not ending the contract. The switch protects your account; it does not erase what you agreed to. Cancel first, keep the proof, then stop the payment as a backstop — and if a real debt is left behind, that is a different set of rights.

Do not wait for Washington. The FTC restarted in March 2026 at the advance-notice stage — the step before a proposal — and comments closed that April. The last attempt took two years to write and never applied to anyone for a single day. The bank rule is in the book this morning.[7, 37]

Frequently Asked Questions

Short answers to the questions that come up most often. Where a full answer needs more room, the relevant section above has it.