Dear Customer, Fuck you.
We've updated our Terms of Service and Privacy Policy. Effective immediately (or whenever we decided, really), we may use your content, code, chats, interactions, telemetry, and whatever else you leave on our platform to train, improve, and develop AI models. We may also share it with third-party AI model providers. By continuing to use the service, you agree. If you don't like it, stop using the service.
A frontier AI lab showed up with a check large enough to make the question "should we sell our users' data?" feel rhetorical. Not "maybe we can anonymize and aggregate." Not "only if users opt in." Just: here is the number. The number at which the internal debate ends and the legal team starts drafting the "continued use constitutes acceptance" language. That number is the new definition of fuck-you money. It is the amount required for a company to accept the lab's terms and then turn around and tell you that if you object, you can fuck off.
This is not rare. It has become routine. Developer platforms, consumer apps, productivity tools, and social services have been quietly rewriting the same clause for a couple of years. Free and lower-tier users are often opted in by default. Paid tiers sometimes get a slightly softer default or a self-serve toggle. Enterprise gets the polished contract language that says "we won't train on your data" while the consumer side of the same company does the opposite. Opt-outs exist in many places, but they are usually account-level, non-retroactive, and full of exceptions for "safety," "feedback," and "already-trained models." Data that already went into a previous training run does not come back out.
The public licensing market gives a rough sense of the scale of the checks involved. Multi-year deals in the tens to low hundreds of millions of dollars for high-value content sources are no longer surprising. Most of the money never reaches the people who actually generated the data. It reaches the platforms that sit between the users and the labs. The ToS update is simply the mechanism that turns your ordinary use of a product into a licensable asset.
The legal form is familiar: notice plus continued use. Clickwrap for the original agreement, then passive acceptance for the expansion. Courts often uphold it when the notice is clear enough. Regulators have warned that quietly expanding prior privacy promises can cross into unfair or deceptive territory, but the updates keep coming. The practical effect is the same in almost every case. The company keeps the data rights. You keep the product, or you leave.
So the email you receive is accurate, if incomplete. We found our price. The lab found our price. You are now being informed of the transaction after the fact, with the traditional customer-service flourish: if you disagree, cancel.

The core issue presented is the mechanism by which platforms are updating their Terms of Service and Privacy Policies to utilize user-generated content, including chats, code, interactions, and telemetry, for the purpose of training, improving, and developing artificial intelligence models, often involving sharing this data with third-party providers. This update is framed within a context where the value of this data is substantial, quantified by large financial figures that represent the acceptance threshold for companies regarding user data usage.

The text argues that this practice of data acquisition and utilization is routine across various sectors, including developer platforms, consumer applications, productivity tools, and social services, where the terms are quietly rewritten over time. While opt-out mechanisms exist in many areas, they are typically limited to account-level settings, are non-retroactive, and include exceptions for reasons such as safety or already-trained models. Furthermore, data from previous training runs is generally not made available for withdrawal.

The structure of this legal agreement relies on a familiar pattern: providing notice followed by continued use, often structured as a clickwrap agreement where passive acceptance is implied. Although regulators have cautioned about the potential for expanding privacy promises to cross into unfair or deceptive territory, the practical outcome remains consistent: the companies retain control over the data rights, and the user must choose between leaving the service or continuing to use the product. The text posits that the Terms of Service update functions as the mechanism that transforms ordinary product usage into a licensable asset.

The scale of this monetization is significant, with multi-year deals amounting to tens to hundreds of millions of dollars for high-value content sources. Critically, the majority of this revenue accrues to the platforms that mediate between the users and the AI labs, rather than the original data generators. This dynamic underscores how the contract language effectively monetizes user input. The text concludes by summarizing the asymmetry of this situation: the user is informed of the transaction only after the fact, facing a requirement to cancel if they disagree, while the platforms maintain the data rights.