What happened
Stripe research found more than one in six sign-ups at AI companies come from multi-account abuse, and Stripe Radar blocked $1.3 billion in fraud in Singapore last year. Lovable moved from subscriptions to usage-based billing.
Why it matters
Every free token an abuser burns is an immediate cost the AI company never recoups, so even cheap compute may not protect margins. The real test is whether usage turns into revenue.
What to watch
Catching abuse at sign-up is the hinge, since acting only after a payment fails is too late. Watch whether AI firms adopt usage-based pricing and risk controls like Stripe Radar.
WHO IT HITSAI company finance and trust-and-safety teams, plus product managers setting pricing, are most exposed: abuse at sign-up turns free-trial tokens into unrecovered operating costs before any payment is attempted.
Summaries like this, every morning.
Software economics long rested on a simple promise: build once, then serve each extra customer at almost no cost. Generative AI breaks that. Every prompt, answer, and agentic task must be produced and paid for again at the moment of use, so made-to-order software carries made-to-order costs. Stripe co-founder Patrick Collison captured the shift by arguing software should be like pizza, made fresh on demand.
The internet economy was built around human intent: people search, click, subscribe, and check out, letting businesses convert a customer and then serve them cheaply. AI agents change that, researching, writing code, and paying on people's behalf, which can lift consumption far past what a business expects. In Asia, adoption is moving from experiment to action: a 2025 Stripe survey found 82% of business leaders were already implementing or planning to implement agentic AI, and Bain reports 85% of Southeast Asian shoppers use or consider AI tools for shopping.
As tokens become the unit that matters, pricing models are shifting toward usage-based or hybrid plans, as Lovable did when it moved beyond subscriptions to charge on token consumption past a plan's free allowance. That openness also invites token theft, in which attackers spin up accounts to burn free tokens. The stakes hinge on early detection at sign-up, before the first token is spent, and on giving companies the visibility and flexibility to treat a determined abuser differently from a paying user whose usage merely spikes.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
