Binance's policy arm published a new paper today arguing that financial compliance needs an agent-specific identity layer before autonomous software starts moving real money at scale. The paper, "Know Your Agent: Building the Compliance Layer for Agentic Finance," comes from Steven McWhirter, Binance's Global Head of Policy, and crypto policy manager Diego Montes Serralde. The pitch is a five-layer framework meant to sit alongside — not replace — existing KYC and AML obligations. The underlying worry is straightforward: traditional compliance was built around knowing the human customer, and that model breaks down once an AI agent is the one initiating transactions on a customer's behalf. Binance frames this as closing an accountability gap rather than inventing new regulation from scratch. This isn't happening in isolation. The paper references Singapore's MAS SAFR framework from July 2026, which focuses on runtime safeguards and auditability for financial AI systems, and notes that the UK's Financial Services AI Adoption Plan has already recommended a 'Know Your Agent' protocol paired with interoperable machine-to-machine authentication. That's two regulators and now a major exchange converging on the same basic idea within a few months of each other. What's notable is who's making this proposal. Binance isn't a neutral standards body — it's an exchange with a direct commercial interest in agentic finance working smoothly and in not becoming the venue for the next AI-driven breach headline. A 'Know Your Agent' protocol, if it takes hold, would function less like a registry (an agent names itself) and more like a bank-grade due-diligence layer: before an agent moves funds, something or someone has to vouch for which entity it's actually acting for, under what authorization, and with what limits. The gap this is trying to fill is real. Agent-to-agent payment specs like Google's Agent Payments Protocol treat individual purchases as signed claims, but that solves authorization at the transaction level, not identity at the account level — who is this agent, who operates it, and does it have standing to act on this account at all. KYA as a concept tries to answer that upstream question, before any single transaction gets signed. Whether exchanges, regulators, and agent platforms converge on one shared KYA standard or ship incompatible versions is the open question — and given how fragmented agent identity specs already are, convergence is not guaranteed. No implementation timeline or technical spec has been published yet; this is a policy paper framing the problem, not a working protocol.