Seattle Times and Newsday sue OpenAI and Microsoft over copyright, paywall bypass and lost revenue

Somebody at a regional paper owns the meter configuration, and this week that person acquires a legal audience. The Seattle Times and Newsday complaint argues about how the articles were reached and not only about what the models produced, which is the distinction TNW drew in reading the filing ^[2][3]. Product decisions sit on that side of the line.

Publishers mostly treat the wall as a revenue instrument: how many free stories a visitor gets before the prompt appears, and which referrers get waved through. The filing treats it as a statement of the terms on which a page may be accessed, and alleges those terms were bypassed methodically ^[2]. Those are two different objects sharing one settings screen.

Microsoft's response, according to TNW, went at the money rather than the mechanism. On the same day, in the same court, it asked for rejection of the two theories the complaint rests on, lost licensing fees and market dilution ^[11]. TNW's reading is that the paywall allegation travels furthest of the four in the complaint, because Microsoft's fair use brief argues about books rather than about how pages were reached ^[19].

In Europe the same conduct is something a company has already promised to avoid. The copyright chapter of the EU's general-purpose AI code commits signatories not to circumvent effective technological measures restricting access, with specific attention to subscription models and paywalls, and it requires crawlers to read and follow robots.txt ^[13][14]. OpenAI signed it ^[12]. The code carries two limits worth knowing before anyone builds a strategy on it. It has applied only since August 2025, roughly two years after the New York Times filing that opened this front, so the training runs at the origin of the litigation sit outside its commitments ^[18]. And whether a paywall counts as a valid reservation of rights is unsettled, because the directive asks for machine-readable means and no court has ruled that a wall qualifies ^[15]. What is settled is that Europe has no fair use equivalent, and that a German court held in July, in the case brought by GEMA, that works memorised in model parameters infringe at the point of storage ^[16][17].

The forcing question for a publisher with a meter has two axes. First, whether the archive is worth licensing to a lab, the route AP and Vox Media took ^[10]. Second, whether you can produce a record of what an unauthenticated request to your article pages actually returned, and when. Licensable with a record, and you negotiate from evidence. Licensable without one, and you are selling reputation. Not licensable but holding a record, and you have the claim that the 400 local newspapers already in this litigation are pressing collectively ^[5]. Neither, and you are a spectator to someone else's discovery.

The cost on that second axis is real, and it is not a legal cost. Whatever the meter gives away to convert casual readers it also gives away to anything that requests the page the same way, and enforcing the wall server-side removes the sampling that sells subscriptions. That tradeoff used to be settled by conversion numbers alone. It now also decides what a publisher can prove about how its pages were reached.