Fragment raises $12M to guard AI agents against payment errors

New York-based Fragment has raised $12 million to build software that double-checks transaction math before payments initiated by AI agents can clear. The round was first reported by MSN.

The startup’s product sits between AI agents and payment rails, verifying calculations and flagging discrepancies before funds leave company accounts. Fragment does not disclose customers, but its pitch targets enterprises adopting agentic AI for routine operations like vendor payments, expense reimbursements, and payroll.

AI agents that handle financial transactions create a new attack surface: errors in logic or data can lead to significant, irreversible transfers. Fragment’s approach treats these agents as high-risk users, imposing real-time audits without breaking their autonomy. The company claims its software can catch mistakes that traditional fraud-detection systems miss because it understands the context of each transaction—whether a payment aligns with expected amounts or patterns.

The round arrives as agentic AI startups attract substantial funding. In the past month, StartupReader has covered multiple rounds for companies building AI agents or the infrastructure they rely on. Fragment’s focus on financial safety suggests a maturing market where businesses are moving from pilot projects to full deployment—and looking for tools to contain the risks.

Sources: msn.com