The Problem: Fraud Happens Before Checkout
On October 7, 2026, FinTech Global published an analysis by Vivox AI from a panel held at QUBE Events' sixth Financial Innovation Forum: Payments and RegTech. The panel, moderated by Bank of London deputy director of financial crime Zainab Shode, included leaders from Mastercard, Starling Bank, PXP, and CGI.
The core finding: AI agents can make payments that pass every check and still be fraudulent. Consider a simple instruction to book a hotel. An agent might pick a convincing fake website and pay a fraudulent merchant with the customer's full authority. In a more extreme scenario, it could cancel another guest's booking to secure the room. In both cases, the final payment reveals only part of the story.
The problem is that traditional fraud controls inspect the transaction at checkout: amount, merchant, card details, location. But when an AI agent is making the decision, fraud can happen upstream. Fake merchant sites, fabricated catalogs, and malicious instructions hidden in content agents read can all manipulate the agent's behavior long before a payment is processed.
Consent Is Not a Single Approval
The panel framed consent as a set of specific, checkable permissions rather than a single approval. Payment providers need to know:
| Permission Dimension | What It Specifies | Why It Matters |
|---|---|---|
| Spending limit | How much the agent may spend | Prevents runaway costs from manipulated instructions |
| Merchant scope | Which merchants or categories are allowed | Blocks payments to fraudulent merchants |
| Time limit | When the agent's authority expires | Limits exposure if the agent is compromised |
| Revocation | How a customer amends or withdraws authority | Ensures changes are recognized across the payment journey |
These rules will vary between cards, digital wallets, and account-to-account transfers, each of which handles authorization, disputes, and liability differently. Revoking an agent's access in an app is only meaningful if that change is recognized across the entire payment journey.
Identity Gets More Complex
Firms must confirm not only who the customer is, but which agent is acting and whether it is the one the customer authorized. Vivox AI highlights the value of records that link customer identity, agent identity, and the scope of authority, allowing investigators to reconstruct what was requested, what the agent encountered, and what it did.
This three-part identity chain, customer to agent to action, is fundamentally different from the two-part chain that traditional payment fraud systems use. It creates a new verification layer that most existing systems were not built to handle.
Nacha Names nsKnox Preferred Partner for B2B Fraud Prevention
On October 8, 2026, Nacha announced that nsKnox has become a Nacha Preferred Partner for Account Validation, Risk and Fraud Prevention, and ISO 20022. This partnership addresses a related but distinct problem: B2B payment fraud, particularly account validation attacks where fraudsters redirect payments to accounts they control.
nsKnox uses AI-powered signals to secure ACH Network transactions, providing real-time account validation that helps prevent supplier impersonation and payment redirection fraud. The partnership signals that the infrastructure layer of payment fraud prevention is evolving to incorporate AI-driven verification, not just transaction monitoring.
The combination of Vivox AI's agentic payments analysis and Nacha's nsKnox partnership illustrates two sides of the same problem. Consumer-facing agentic payments need new consent and identity models. B2B payments need stronger account validation. Both need controls that can distinguish legitimate automated activity from misuse.
Fraud Exploits Speed and Repetition
Once attackers find a weakness in agentic payment systems, automated agents could exploit it repeatedly and at speed. The panel pointed to several attack vectors:
- Fake merchant sites: Agents that scrape the web for deals may land on fraudulent storefronts designed to look legitimate.
- Fabricated catalogs: Product listings manipulated to redirect agents toward fraudulent purchases.
- Malicious instructions: Content that agents read during research, such as compromised reviews or injected prompts, can steer decisions.
- Permission interference: Attacks that target the moment when permissions are first set, exploiting gaps in the consent flow.
The speed at which agents operate means that a vulnerability discovered at 9 AM could be exploited thousands of times by noon. Human-in-the-loop checkpoints become essential for high-value transactions, but they cannot be the only safeguard if agents are to operate autonomously.
Liability Remains Unresolved
Panel participants disagreed on how responsibility should be shared between customers and the firms supplying agents, payments, and safeguards. They also disagreed on how existing reimbursement rules would apply to agentic transactions. What they agreed on was the need for evidence detailed enough to pinpoint where a failure occurred.
This unresolved liability landscape is a significant risk for businesses. Until regulatory frameworks catch up with agentic payments, companies deploying or enabling AI agent transactions need to define their own liability policies and ensure their terms of service address agent-initiated payments.
What Businesses Should Do Now
Vivox AI's takeaway is clear: firms should test now whether their fraud, money laundering, and sanctions controls can separate legitimate automated activity from misuse, rather than waiting for a finished monitoring model. Practical steps include:
- Map your agent payment surface. Identify every point where an AI agent could initiate, modify, or authorize a payment.
- Test existing controls against agentic scenarios. Run simulated attacks where an agent is manipulated upstream and see if your checkout-level controls catch it.
- Implement scoped consent. Move from single-approval to permission-based models with spending limits, merchant restrictions, and time boundaries.
- Build the identity chain. Ensure your systems can link customer identity, agent identity, and action scope for every transaction.
- Layer account validation. For B2B payments, adopt solutions like nsKnox that provide real-time account validation against redirection fraud.
- Establish human checkpoints. For high-value or unusual transactions, require human confirmation before the agent proceeds.
Read Next
- The 2026 AI Trust Report: Which Brands AI Actually Trusts — 3.9 million citations reveal AI trust patterns
- How QuickBooks Earned a 223% Citation Lift in AI Search — A concrete GEO playbook
- Google's Gemini Coworker Agent Gets Its Own Email — Persistent AI identity for enterprise work
- Automation Anywhere Acquires Boost.ai: The Conversation-to-Outcome Loop — The autonomous enterprise vision
