March 2026 Summaries

Stripe recently identified a significant increase in first-party fraud, particularly in the form of free trial abuse, through an analysis of transactions across its network. This trend is notably prevalent among AI companies that are vulnerable due to their reliance on costly compute resources and free trials to attract customers. Bad actors exploit these trials without transitioning to paid subscriptions, resulting in substantial financial losses, especially for startups offering self-service signups and direct API access. This pattern extends beyond AI firms to various industries such as SaaS platforms and marketplaces. To combat this, Stripe has enhanced its AI-powered fraud tool, Radar, to detect and prevent free trial abuse with high accuracy by analyzing payment instruments and behaviors. This tool has already proven effective, blocking over 550,000 high-risk trials and averting an estimated $4.4 million in losses for several AI businesses. Radar's solution is available for all industries, and interested parties can request early access through Stripe's contact channels.

At the Merchant Risk Council (MRC) Vegas 2026 conference, over 2,000 payments leaders discussed evolving fraud patterns, highlighting the need for dynamic, tailored interventions to counter increasingly automated fraud. Traditional uniform authentication methods are being replaced with strategies that assess user intent and employ high-trust velocity, as exemplified by companies like Stripe and Airbnb. These strategies aim to reduce friction for trusted users while introducing multilayered identity verification to combat deepfakes and synthetic identities. Additionally, the rise of agentic commerce is pushing enterprises like Ashley Furniture to integrate fraud detection directly into payment infrastructures, using real-time adaptive techniques that overcome the limitations of static, rules-based systems. This evolution in fraud prevention is driven by the need to address sophisticated fraud tactics, such as those involving deepfakes, which now require a comprehensive approach to identity verification to detect anomalies that bad actors can't perfectly replicate every time.

Subscription businesses are increasingly global, with AI-driven growth necessitating complex pricing strategies across diverse markets. Localizing prices for one-time purchases is already challenging due to fluctuating exchange rates and the need to manage various currencies, but subscriptions add further complexity as prices must remain consistent across billing cycles. To address these challenges, Adaptive Pricing, part of the Optimized Checkout Suite, allows businesses to present prices in local currencies while Stripe manages currency conversion, ensuring price stability despite exchange rate changes. This approach enhances conversion rates and subscription lifetime value (LTV), as demonstrated by a study of 1.5 million subscription sessions that showed a 4.7% average increase in conversion and a 5.4% increase in LTV. Localized pricing not only improves customer understanding and trust but also increases retention rates, as seen in examples where businesses like Runway experienced significant LTV gains. By adopting Adaptive Pricing, businesses can scale globally without needing to develop their own foreign exchange infrastructure, thereby capturing more revenue and improving overall subscription performance.

AI is transitioning from basic chatbots to advanced autonomous agents capable of planning, executing actions, and evaluating outcomes, highlighting the need for a financial system that accommodates these agents. Current financial tools are human-centric, creating challenges for agents in transactions, such as account creation and payment processing, which often require human involvement. To address this, the Machine Payments Protocol (MPP), developed by Tempo and Stripe, offers an open standard for agents to manage payments seamlessly, supporting microtransactions and recurring payments in stablecoins and fiat currency. MPP is integrated with Stripe's infrastructure, allowing businesses to accept agent payments using Shared Payment Tokens (SPTs) and enabling new business models, like Browserbase's session payments and Prospect Butcher Co.'s autonomous sandwich orders. MPP ensures that agent transactions are processed like human ones, with benefits like fraud protection and tax calculations. Stripe's Agentic Commerce Suite and Protocol aim to expand this agent economy, offering comprehensive financial tools for developers and businesses to engage with this emerging user category.

Agentic commerce, the integration of AI agents in purchasing processes, presents significant challenges and opportunities for businesses, as demonstrated by the launch of the Agentic Commerce Protocol (ACP) and the Agentic Commerce Suite by Stripe and OpenAI. This new commerce model requires sellers to adapt their product catalogs for different AI agent formats, manage real-time inventory verification, and address the complexities of agentic transactions, including catalog discovery and fraud prevention. Stripe's solutions facilitate these processes by providing a protocol-agnostic commerce layer, handling Shared Payment Tokens (SPTs), and integrating fraud protection through its network, ensuring secure and efficient transactions. As agentic commerce evolves, businesses are advised to strategically select SKUs, payment methods, and fulfillment options to optimize this new channel's potential, while also considering whether to develop first-party agent experiences or leverage third-party AI agents for broader market reach. With advancements in machine payments and global expansion plans, Stripe aims to streamline agentic commerce and enhance sellers' ability to engage with AI-driven consumer interactions.

From November 2025 to February 2026, Stripe observed a rise in first-party fraud, where legitimate users exploit policies like free trials and refunds, significantly impacting businesses and costing $35 for every $100 in disputes. This fraud manifests in three main forms: account abuse during sign-up, free trial abuse during product evaluation, and refund fraud post-purchase. AI companies face particular vulnerability due to the high compute costs associated with multiaccount abuse, with 7.4% of AI customer sign-ups implicated in such fraud. Traditional methods to mitigate these abuses, such as blocking virtual cards, are proving ineffective, especially for AI firms that rely on free trials for customer acquisition. Refund abuse further costs businesses worldwide approximately $100 billion annually, with lenient return policies being exploited by consumers and influencers alike. Stripe is developing new features in its AI-powered Radar product to help businesses identify and combat these forms of first-party fraud, offering early access to tools that predict and block suspicious activities, ultimately aiming to provide comprehensive fraud protection across its network.

Stripe's launch of Shared Payment Tokens (SPTs) has revolutionized agentic commerce by allowing agents to initiate payments using a customer's preferred method without exposing sensitive credentials. This initiative has seen significant adoption by major companies like Etsy and URBN, prompting Stripe to expand SPT support to include popular payment options such as Mastercard Agent Pay, Visa Intelligent Commerce, and buy now, pay later (BNPL) services like Affirm and Klarna. This makes Stripe the first provider to integrate both agentic network and BNPL tokens through a single platform. Mastercard and Visa have collaborated with Stripe to enable secure, AI-driven payments using network-issued digital credentials, which allow authorized agents to make purchases on behalf of customers without exposing card details. Stripe simplifies the process for sellers by managing the complexities of token provisioning, allowing them to seamlessly support new payment methods. The integration of BNPL options has also driven increased revenue for businesses by boosting conversion rates and average order values. As Stripe continues to innovate in agentic commerce, it plans to broaden SPT support to include more payment methods, enhancing accessibility for customers.

State-of-the-art language models (LLMs) can now solve a majority of scoped coding problems, but there remains a gap between this ability and the capacity to autonomously manage entire software engineering projects, particularly in real-world settings that require complex integrations such as those with the Stripe API. A research team developed the Stripe integration benchmark to assess the ability of LLMs to handle backend, full-stack, and specific Stripe feature tasks, revealing both the strengths and limitations of these models. Surprisingly, LLMs like Claude Opus 4.5 and OpenAI’s GPT-5.2 showed proficiency in full-stack tasks and API feature sets, respectively, but still struggled with handling ambiguous situations and completing browser-based tasks. The benchmark was designed to push the models’ capabilities by creating complex environments that mimic realistic software development scenarios, and it serves as a testing ground for improving agentic tools through iterative evaluations. The project highlights the continued need for rigorous testing and refinement of LLMs to ensure fully accurate and autonomous software integrations, inviting collaboration and feedback from the broader software community to advance these developments.

Stripe has implemented a series of updates and enhancements across its platform over the years, focusing on improving user experience, expanding global reach, and introducing new functionalities. Recent developments include enabling Connect platforms to embed account onboarding and payment components, supporting automated US sales tax filing via TaxJar, and introducing the Accounts v2 API for better onboarding. Stripe has expanded its payment options, such as offering PayTo direct debits in Australia and supporting various "buy now, pay later" methods like Klarna and Alma in multiple regions. The platform has also enhanced its analytics and data management capabilities with tools like the Payment Records API and Stripe Sigma's new tax transaction tables. Additionally, Stripe has improved its billing features with integrations for multiple payment processors, advanced usage analytics, and flexible pricing models. Significant efforts have been made to enhance fraud prevention through Radar's dynamic risk controls and custom rule-writing capabilities. Stripe continues to support global business operations by introducing new payment methods, tax compliance tools, and financial account management features in various countries.