Socure launched Agent Intelligence on Tuesday, October 27, 2026, a suite that detects AI agents visiting a website, scores whether they should be trusted, and ties each agent to the verified person or business that authorized it, answering a question that is getting harder for every online business: when an AI agent shows up at your digital door, can you trust it? The launch arrives as agent traffic surges, with Socure reporting that daily requests from AI agents grew by more than 1,700 percent in a year to September 2026, and that for the first time more than half of internet traffic was not human.
Three Layers, One Identity Graph
Agent Intelligence is built as three layers on Socure’s RiskOS platform, designed to go beyond simple “know your agent” checks. The first layer, Agent Detection, is generally available now inside Socure’s existing Digital Intelligence offering. It identifies and classifies AI agents driving browser sessions, whether or not they declare themselves, and names the agent when it can be identified. Because control of a session can pass between a human and an agent mid-session, it returns a confidence score and reason codes rather than a binary verdict.
In controlled lab testing, Socure said named consumer agents such as Muse and Instinct were identified with 99.99 percent accuracy across 700 behavioral variables. That is a lab result, not a field guarantee, and customers will judge it by false-positive rates on real traffic, not by lab scores.
The second layer, Agent TrustScore, is open for beta customers now. It is a composite score built from three data categories, returned separately so customers can weight them for their own use case: identity, meaning who the agent is and who stands behind it; authorization, meaning what the agent is permitted to do and whether the action falls within that permission; and risk, meaning what the histories of the principal and operator indicate, including signs of replay or impersonation.
The Credential That Arrives Next Year
The third layer, Agent Credential, reaches general availability in the first quarter of 2027. It is a W3C Verifiable Credential, issued through SocurePass, that operators give to their agents, binding each agent to a verified principal in Socure’s identity graph. Any site can verify the credential without calling Socure, and because the format supports cryptographic chaining and proof of possession, integrity can be carried through a delegation chain when one agent hands work to another.
Socure deliberately made Agent Intelligence agnostic to agent type, presentation, and credential issuer, so TrustScore evaluates credentials from any issuer as well as agents that carry no credential at all.
Block Them All, Or Let Them All In
Socure’s pitch is framed as a third option between two failing strategies. “Block them all and you block your next customer. Let them all in and you hand fraudsters the keys, with little to no controls,” said Johnny Ayers, Socure’s founder and chief executive. The company argues that most organizations still treat agents like bots and block them, a strategy that stops working the moment a best customer tries to engage through an agent such as Muse or Instinct.
The commercial logic is easy to see. According to Accenture’s 2026 Consumer Pulse report, cited by Socure, 85 percent of consumers were open to collaborating with an AI agent and 74 percent were open to having the agent execute specific tasks. If those consumers start shopping, booking, and banking through agents, businesses that cannot tell a legitimate agent from a fraudster’s script face an impossible choice between lost sales and open fraud.
Why It Matters
This matters because the agent economy breaks the internet’s oldest assumption: that a session belongs to a person sitting at a keyboard. When agents act for people, every verification method built for humans, from passwords to captchas, becomes a weaker signal, and the question shifts from “is this a human?” to “whose agent is this, what is it allowed to do, and can I prove it?” Socure’s answer, detect, score, and credential on one graph, is coherent, but its real test is adoption: whether agent operators bother to issue credentials when they ship in 2027, whether relying sites actually check them, and how regulators and courts assign liability when a trusted agent does something its owner never asked for. Giving agents verified identities and controls is a race several vendors are now running, and tools built specifically for AI agents are trying to give them a proper place to work rather than a patchwork of scripts.
Background
Socure describes itself as an AI-native trust infrastructure company for global identity and risk intelligence, serving more than 6,000 customers in 190 countries across financial services, government, gaming, health care, telecom, social media, marketplaces, technology, and e-commerce. Its customer base includes the top five United States banks, four of the Magnificent Seven technology companies, and more than 700 fintechs. TrustScore beta and the 2027 credential rollout will show whether the market wants agent trust sold as a suite or assembled from separate parts. Enterprise AI is meanwhile moving agents deeper inside business systems, which only raises the stakes for knowing whose agent is acting.
0 Comments