LexisNexis Risk Solutions Adds AI-Agent Signals to Fraud Decisions

LexisNexis Risk Solutions announced Know Your Agent (KYA) intelligence on November 5, 2026, adding AI-agent activity and contextual risk signals to its digital identity and fraud-prevention platform. The company said agentic traffic grew more than 750% in the first half of 2026.

Many customers already use KYA in current risk-management decisions. The company reported that its Digital Identity Network analyzed more than 116 billion digital transactions in 2025 and detected 4.9 billion automated bot attacks, 59% more than the prior year. That existing visibility into human and automated interactions provides the context into which KYA adds agent activity.

Detection Feeds a Broader Risk Assessment

Most personal agents do not declare themselves as agents, making detection the immediate challenge. Identification alone cannot establish whom an agent represents, whether its activity is authorized, whether its behavior matches that authority or whether the wider interaction appears legitimate. KYA is designed to add agent detection to identity, device, network, behavioral and relationship context in real time.

That context can change during a single digital journey when control passes between a person and an agent. Mike Woodberry, senior vice president of product and analytics, said the goal is to establish what is acting, who or what it represents, whether it is authorized and what it is attempting to do. Each decision therefore concerns a specific agent performing a specific action in a specific context, governed by the organization’s risk policies.

How the Signals Reach Fraud Workflows

The company’s agentic-intelligence brochure describes browser, device, operating-system and network intelligence as ways to recognize agent signatures and other indicators of agent-driven activity. Applications embedded within ChatGPT, Claude and similar platforms can also be profiled through browser characteristics, domains, application signals and other indicators to establish digital-identity context. Technical and behavioral indicators help classify activity as human, malicious bot or AI agent.

Trust Tags can establish persistent relationships among users, devices and agent interactions, helping organizations assess later activity. Out-of-band authentication can help establish trusted relationships among customers, agents and interactions. Organizations can then incorporate the signals into existing workflows and allow, authenticate, review or reject an interaction according to its context and their own risk policies.

Within ThreatMetrix, agentic intelligence combines digital identity and behavioral intelligence. APIs, software development kits and workflow-ready integrations support web, mobile and backend environments, while delivery through the LexisNexis Dynamic Decision Platform connects the capability to forensics investigations, case management, reporting and workflow orchestration.

Fraud Uses and Product Boundaries

Application areas include agent-assisted commerce, account opening and onboarding, banking and payments, authentication and access management, and customer service. Agents may research or buy products, participate in registration journeys, access accounts, schedule returns or manage accounts. The system is also intended to assess whether an agent has legitimate authority during account access and verification.

LexisNexis cautions that autonomous agents could increase the scale and sophistication of authorized push payment fraud and account-takeover attacks, with malicious agents learning from successes and failures to tailor subsequent attacks. Chief commercial officer Grayson Clarke said a global financial institution used the company’s detection capabilities to identify agentic activity and help prevent it from executing sensitive transactions.

Current capabilities center on detection, context and workflow decisions. The materials describe more sophisticated agent identity and authorization, along with agent-to-agent interactions, as future directions without committing to functionality. Results vary with implementation, data availability, transaction characteristics and customer use, and no solution can identify, classify, authenticate or detect every user, device, bot or AI agent in every circumstance.

For fraud teams, the practical change is that agent involvement becomes another input to an established decision, with responses calibrated to the customer, authority and action involved. The open question is whether that contextual approach remains effective as agents, standards and agent-to-agent interactions develop, particularly while the more advanced identity and authorization capabilities are still future directions.