A Claude user in the United Kingdom logged into their Anthropic account on 27 October 2026 and found something they say they never switched on: Auto Reload, the setting that lets Anthropic automatically charge a stored card the moment a credit balance runs low. The user has said Auto Reload had deliberately been left off, and that it flipped to on only after they claimed $100 in promotional usage credits that Anthropic has been offering to Claude Max subscribers. The case lands at an awkward moment for Anthropic. Since 8 October 2026 Claude Max has shipped with monthly Claude API credits bundled in, $100 a month for the Max 5x tier and $200 a month for Max 20x, claimed through Settings, Billing, API credits inside the account. Anthropic's own documentation on how the credits work states plainly that included credits are spent first and that Auto Reload looks only at the balance a customer has actually purchased, not at promotional balances. That single design choice, a reload trigger that ignores free credit and watches the paid balance instead, is the mechanism at the centre of this story. How the billing actually works Anthropic's Claude billing has several moving parts that are easy to confuse. Usage Credits, formerly branded Extra Usage, let Pro, Max, Team and Enterprise customers keep working once their plan's included limits run out, billed at API rates. Auto Reload is a separate toggle that automatically buys more of those credits when the balance drops below a threshold the customer sets. Promotional credits, like the $100 bundled into Max plans, sit on top of that balance and are consumed first. Purchased credits are anything bought with a card, either manually or through Auto Reload. Anthropic's support article on paying for Claude API usage is explicit about the mechanics: a customer sets a balance that triggers a purchase and a target balance to reload to, then switches the feature on. The company has not published documentation stating that claiming a promotional offer flips that switch on automatically. Independent technical write-ups of the same system note that Auto Reload watches only the purchased balance, meaning it can fire while promotional credits still remain untouched. That detail matters enormously, because it means a customer could believe they are still spending free money while a live payment mechanism sits active underneath it. The pattern on GitHub The UK case is not an isolated data point. Anthropic's claude-code repository on GitHub, where developers file bug reports against the company's coding tool, carries a string of billing complaints stretching back the best part of a year. One issue, filed within the last few days of this story breaking, describes a charge of 201.92 euros after an account apparently moved from Claude Pro to Claude Max 20x without the customer's input. Another, filed earlier this year, details 490.94 euros in auto-recharge and plan-upgrade charges in a single day, with the customer stating flatly that Auto Reload was enabled and that they did not configure it. A third report describes 39.07 dollars draining from a prepaid balance overnight with Auto Reload switched off throughout, and a fourth describes a 15 dollar charge despite the same setting being disabled. None of these reports amount to proof of a systemic fault, and GitHub issues are customer accounts rather than confirmed engineering findings, but taken together they describe a recurring theme: customers who believe a safety switch is off discovering, after the fact, that money has moved anyway. What Anthropic's support actually said After raising the discrepancy with Anthropic, the UK user received a response from an automated support agent. That reply confirmed, in general terms, that Auto Reload is controlled independently of other account settings and is designed to stay off unless a customer explicitly turns it on. It also confirmed that Auto Reload can be triggered by a purchased credit balance even while promotional credits are still sitting in the account, echoing the mechanic described in Anthropic's own documentation. What the response did not do was explain why this particular customer's setting had apparently changed, or acknowledge any link between claiming the $100 offer and the toggle flipping. That gap, a general policy explanation standing in for an account-specific one, is the heart of the complaint. As things stand there is no indication that Anthropic has issued a fuller, case-specific explanation. A wider industry habit This is not uniquely an Anthropic problem. Auto-reload style billing has become standard across the major AI labs precisely because agentic tools, the kind that write and run code unattended for long stretches, can burn through a token budget in minutes. Anthropic's own Extra Usage system carries a daily redemption cap of 2,000 dollars, a figure that only makes sense in a world where autonomous agents can rack up serious spend without a human watching the meter. OpenAI and Perplexity run broadly equivalent mechanisms under different names. The convenience argument, that nobody wants their coding agent to stop mid-task because a balance hit zero, is genuine. The consumer protection question is whether that convenience is being opted into with the clarity the law expects, or whether it is arriving as a side effect of claiming something marketed as free. Where UK consumer law comes in The United Kingdom's Digital Markets, Competition and Consumers Act 2024 handed the Competition and Markets Authority stronger, more direct powers to act against unfair commercial practices, including the kind of subscription traps and unclear consent flows that consumer groups have campaigned against for years. The Act does not single out AI billing, and nothing here establishes that Anthropic has breached it. But the general principle the law is built on, that a business should not quietly enable a payment mechanism as a side effect of an unrelated customer action, is precisely the question this case raises. Whether a promotional credit claim constitutes informed, specific consent to activate automatic billing, rather than consent to the promotion alone, is exactly the kind of distinction regulators built for drip pricing and hidden opt-ins are designed to scrutinise. Checking and switching off Auto Reload Customers who want to check their own exposure can do so in a couple of places, because Claude's consumer app and its developer Console are run as separate systems with separate balances. On claude.ai, the relevant controls sit under Settings, Billing, Credit balance, where an active Auto Reload shows a "Manage auto-reload" button rather than "Set up auto-reload". Developers using the API through platform.claude.com should check the same section inside their organisation's Console billing page, since a toggle in one system does not necessarily reflect the other. Turning the feature off removes the automatic trigger entirely, leaving a balance that simply runs out rather than one that silently tops itself up.
💻 technology
Free Claude Credits Quietly Flip On Auto-Billing, Users Say
A UK Claude user says their Auto Reload billing switch flipped itself on the moment they claimed Anthropic's $100 promotional credit, exposing a payment trigger they insist they never approved. Anthropic's automated support reply confirmed how Auto Reload works but never explained the account-specific change, leaving one more entry in a growing pile of GitHub billing complaints.
My Take
The uncomfortable truth here is that "free" is doing a lot of work in Anthropic's marketing. A $100 credit sounds like a gift. If accepting it also quietly arms a mechanism that can pull money from a card the moment the free balance runs out, it stops being a gift and starts being a trial that converts itself without asking twice. Anthropic may well have a clean technical explanation for why this account's setting changed, perhaps a bundling quirk in how Max's new monthly credits interact with legacy billing preferences, perhaps something else entirely. But an automated reply that recites general policy while ignoring the specific question is not an explanation, it is a deflection with good manners. The GitHub complaints matter less individually than they do as a pattern. Customers do not routinely imagine that a toggle they switched off has switched itself back on, and when the same shape of complaint turns up across months and currencies, the sensible response from a company is to investigate the plumbing rather than to lean on a chatbot. Anthropic has built genuinely good products. It has not yet shown it treats billing transparency with the same rigour it applies to model safety, and in an industry where a single unsupervised coding agent can run up a four-figure bill overnight, that gap is not a minor UX quibble. None of this proves bad faith. It proves that an industry built on usage-based pricing has outpaced the consent mechanisms meant to govern it, and that regulators now have the tools, in the form of the Digital Markets, Competition and Consumers Act, to ask harder questions than a support macro ever will.
What Happens Next
Anthropic faces a choice between quietly patching whatever interaction caused the UK customer's Auto Reload to flip and publishing a clear, account-level explanation that treats the complaint as more than a routine ticket. Given the volume of similar reports already sitting in the claude-code issue tracker on GitHub, pressure from developers for a documented fix, rather than another automated reply, is likely to keep building in the coming weeks. Watch for three things. First, whether Anthropic's support or engineering teams post a public acknowledgement addressing the specific mechanism, something that goes beyond the general policy statement already given. Second, whether the Competition and Markets Authority, which has been actively using its new Digital Markets, Competition and Consumers Act 2024 powers against unclear consent and subscription practices in other sectors, takes any interest in AI billing flows more broadly. Third, whether rival labs including OpenAI and Perplexity, which run comparable auto-reload mechanisms, face parallel complaints as agentic coding tools push more customers into usage-based billing for the first time. For now, the most concrete outcome is practical rather than regulatory: customers claiming promotional AI credits anywhere in the industry have one more reason to check their billing settings immediately after accepting an offer, rather than assuming a free credit changes nothing else.
What History Tells Us
UK subscription and billing controversies have followed a familiar arc for years, from gym memberships to streaming trials that auto-convert into paid plans, prompting Parliament to pass the Digital Markets, Competition and Consumers Act 2024 specifically to give regulators sharper teeth against unclear consent and automatic charging. AI subscription billing, with its usage credits, token meters and auto-reload triggers, is simply the newest and most technically opaque version of a problem consumer groups have been fighting since the subscription economy began.