ABSTRACT
Between October 2024 and July 2026, the leading commercial providers of generative artificial intelligence repeatedly modified the pricing, usage limits, and access terms of subscription services after consumers had already paid for them. This paper documents that pattern through a rapid multiple-case documentary analysis of four providers (Anthropic, OpenAI, Microsoft, and Moonshot AI), drawing exclusively on dated public sources: official announcements, technology press coverage, community bug trackers, and the public record of two formal legal actions. Seventeen documented episodes are organized into a chronology and classified with a two-dimensional framework defined by the trigger of the change (exogenous shocks of capacity, regulation, or safety versus endogenous commercial strategy) and the quality of its communication (managed transitions versus unmanaged or repeatedly revised promises). Two findings emerge. First, post-purchase modification is not an incident but an industry-wide operating pattern, and it moves in both directions: restrictions and reversals, price increases and price cuts. Second, formal accountability (a proposed class action against Anthropic over advertised Claude Max usage multipliers and a regulatory enforcement action against Microsoft over Copilot-linked price increases) concentrates in one quadrant of the framework: endogenous changes communicated in unmanaged or shifting ways. The paper argues that this instability is a structural consequence of selling elastic promises against inelastic compute, discusses implications for consumer protection in AI markets, and proposes an agenda for longitudinal, archive-based tracking of subscription terms.
Gonzalez, Jose David, Moving Targets: Promissory Instability in the Pricing and Access Terms of Generative AI Subscription Services (2024-2026) (July 22, 2026).
Leave a Reply