Stripe has expanded its Radar fraud detection platform to address token theft and free trial abuse at AI-service scale, after finding that one in six attempted sign-ups across AI businesses on its network is a bad actor.
The expansion, announced at Sessions 2026 on April 29, introduces real-time sign-up and usage evaluation specifically tuned for AI-native fraud patterns. Stripe reports blocking 3.3 million risky sign-ups monthly for just eight AI businesses, while noting that free trial abuse has doubled in the past six months.
The threat model is distinct from traditional payment fraud. Bad actors are not stealing credit cards; they are creating synthetic identities to consume AI compute resources, steal API tokens, and abuse free-tier allocations at industrial scale. Credential stuffing, synthetic identity creation, and token harvesting now represent the primary fraud vectors for AI-native businesses.
For finance leaders at AI companies, the 1-in-6 statistic demands a recalibration of fraud budgets. Pre-AI fraud baselines are meaningless when nearly 17 percent of sign-ups are adversarial. The broader Stripe Sessions 2026 infrastructure push positions Radar as the detection layer that makes agent-mediated commerce viable, intercepting bad actors before they access the system rather than after they transact.
Source: Stripe Newsroom
