TL;DR: Agentic commerce is creating a “signal blackout” for traditional fraud controls because AI agents can shop from data centres without the IP, device, or cookie signals merchants rely on, while consumer trust and fraud patterns vary sharply by age group, according to Signifyd. The governance challenge is no longer only fraud detection, but deciding how to authorise, inspect, and differentiate legitimate delegated activity from automation abuse.
At a glance
What this is: Signifyd argues that agentic commerce is breaking traditional fraud signal assumptions by moving shopping activity into AI-mediated flows that no longer present familiar device and network markers.
Why it matters: Fraud, IAM and risk teams need to understand how delegated AI activity changes trust, authorisation and detection because controls built around human browsing patterns will miss or misclassify the new transaction path.
Context
Agentic commerce changes the fraud problem because the actor initiating the shopping journey is no longer always a human on a normal device, network and browser stack. When that activity moves into data centres and other automated environments, long-standing detection models lose the signals they were built to inspect.
For identity and fraud practitioners, the governance question shifts from only detecting bad transactions to deciding how delegated activity should be authorised, differentiated and monitored. The article sits at the intersection of fraud prevention, identity trust and automation governance, which makes the signal gap a programme design issue rather than a tuning problem.
Key questions
Q: What breaks when fraud controls are used on agentic commerce without delegation evidence?
A: Teams lose the signals those controls were built around, so legitimate agent-driven purchases can be blocked while fraudulent automation still slips through. The failure is not simply weaker detection, but a mismatch between human-browsing assumptions and delegated machine activity. Fraud teams need proof of authority, provenance and scope, not just device and network markers.
A: They should prioritise delegated authority whenever shopping can be initiated by an AI agent from cloud infrastructure, because device fingerprinting may be unavailable or misleading. The decisive question becomes whether the agent was allowed to act for that customer and for that transaction, not whether the browser looked familiar.
Q: What are the signs that agentic commerce fraud controls are failing?
A: A strong warning sign is a rising mix of false declines, unexplained approvals and disputes where the transaction path cannot be reconstructed from existing logs. If analysts cannot tell whether a purchase was human-led or agent-led, the control environment is blind to a new class of traffic rather than simply underperforming.
Q: How should fraud and identity teams govern AI agent purchases and disputed orders?
A: They should define policy for delegated transactions, preserve provenance end to end and align fraud review with identity evidence rather than browsing artefacts alone. That governance model helps separate authorised automation from abuse and gives investigators enough context to resolve disputes consistently.
Technical breakdown
Why agentic commerce creates a signal blackout
Traditional fraud systems infer trust from stable behavioural and device-linked signals such as IP reputation, cookies and device fingerprinting. Agentic commerce breaks that assumption because the agent may act from cloud infrastructure rather than a user endpoint, so the merchant no longer sees the same context that usually supports risk scoring. That does not make the transaction malicious by default, but it does make the old evidence set incomplete. The practical challenge is that a weaker signal environment increases both false negatives and false positives when controls were designed around human browsing patterns.
Practical implication: fraud teams need detection models that evaluate delegated intent and transaction context, not just browser and device markers.
Delegate tokens and commerce protocols change the trust model
The article points to emerging standards such as the Agentic Commerce Protocol and Universal Commerce Protocol, which are intended to let agents present delegation information to merchants. In governance terms, that moves the control point from passive identification to explicit delegation handling. A merchant needs to know not only that an actor is automated, but also whether the automation is authorised, scoped and revocable. This is where identity and transaction security overlap: the trust decision becomes a question of entitlement, not simply fingerprinting.
Practical implication: teams should define what evidence constitutes valid delegated authority before accepting agent-driven checkout flows.
Payment-initiation and friendly-fraud risk rises when automation scales
As agents take a larger role in shopping, the article expects more payment-initiation fraud and friendly fraud. That matters because automation can compress the time between discovery, cart building and checkout, leaving less opportunity for manual inspection while also making disputes harder to interpret later. The fraud control problem is therefore not only prevention at the edge, but also post-transaction explainability. Merchants need to distinguish legitimate delegated purchases from abuse patterns that look operationally similar once the human is abstracted away.
Practical implication: build dispute and review processes that retain delegation evidence and transaction provenance from the start.
NHI Mgmt Group analysis
Agentic commerce creates a governance gap, not just a detection gap. Fraud teams have long optimised around visible human behaviour, but delegated shopping changes the identity of the actor at the point of decision. When the buyer may be an agent, the control question becomes who is authorised to transact and under what delegation scope. Practitioners should treat authorisation evidence as part of fraud prevention, not as a separate identity concern.
Signal blackout is the right concept for this market shift. Once shopping moves into data centres and browserless automation, merchants lose the stable signals that made IP reputation, cookies and device fingerprinting useful. That does not mean the control stack is obsolete, but it does mean the primary evidence shifts toward protocol-level delegation, transaction context and provenance. Practitioners should redesign scoring models around what remains observable, not what used to be.
Agentic commerce will force fraud teams to become policy teams. The future operating model is less about simply blocking suspicious orders and more about defining which agent actions are allowed, how they are represented and when they can be revoked. That makes commerce identity governance part of the fraud function, especially where delegated activity may later be disputed. Practitioners should build policy around intent, authority and recoverability.
The rise of friendly fraud in agentic flows will test attribution. When a purchase is initiated by an agent, post-transaction dispute handling becomes more difficult because the trace between user intent and machine execution is less obvious. That increases the value of clean provenance, consistent logs and shared signals across fraud, payments and customer support. Practitioners should preserve evidence of delegation from initiation through settlement.
From our research library:
- Nearly 60% of companies reported that fraud losses were still increasing in 2025.
What this signals
Agentic commerce will push fraud governance upstream. Once shopping is initiated by delegated automation, the key decision is no longer only whether an order looks risky, but whether the agent had a valid path to act at all. That is a trust and authorisation question, so fraud programmes will need closer alignment with identity governance and transaction policy.
Clean provenance becomes the control variable that matters most. If merchants cannot reconstruct how an order was formed, what authority the agent had and where the handoff occurred, post-transaction review will stay expensive and inconclusive. Practitioners should expect evidence retention to become a core requirement for both dispute handling and risk tuning.
Fraud teams that still depend on browser-era trust signals will overblock or under-detect. The operational response is to separate ordinary consumer traffic from delegated machine traffic and to build decisioning around intent, scope and recoverability.
For practitioners
- Define delegated transaction policy Decide which agent-driven shopping and checkout actions are permitted, what authority must be presented, and how revocation works when a session or token is no longer valid.
- Instrument agent provenance in fraud models Add protocol-level and provenance fields to decisioning so models can distinguish authorised agent activity from unauthorised automation when device and cookie signals are absent.
- Retain evidence for dispute handling Store delegation context, checkout provenance and decision timestamps so fraud analysts and customer support can reconstruct why a transaction was accepted or declined.
- Separate fraud review from human browsing assumptions Tune rules for agent-mediated purchases separately from ordinary card-not-present traffic, because the absence of consumer endpoint signals is now expected in some flows.
Key takeaways
- Agentic commerce is exposing a trust gap because merchants no longer receive the stable device and network markers their fraud systems were built to read.
- The practical risk is twofold: good orders can be blocked, and new fraud patterns can hide inside legitimate-looking delegated activity.
- Fraud and identity teams need policy, provenance and delegation evidence to govern AI-mediated checkout rather than relying on browser-era signals.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10 addresses the attack and risk surface, while NIST SP 800-63 and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agent-driven checkout depends on delegated authority that can be abused or misread. |
| ASI09 — Human-Agent Trust Exploitation | The article centres on fraud and trust decisions when humans delegate shopping tasks to AI agents. | |
| Recommendation — Control agent delegation boundaries so checkout actions cannot exceed granted identity and privilege scope. Validate human intent and agent authority separately before approving commerce actions. | ||
| NIST SP 800-63 | SP 800-63C — Federation | Delegated commerce relies on transferable trust assertions and tokens between parties. |
| Recommendation — Use federation-grade assertions to represent delegated authority in commerce workflows. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Fraud decisions now depend on whether an agent is authorised to act for a customer. |
| ID.RA-01 — Risk Management Strategy | The article describes a new operating risk that existing fraud models do not fully cover. | |
| Recommendation — Tie transaction approval to explicit entitlements for agent-mediated purchasing paths. Update fraud risk strategy to treat agentic commerce as a distinct decisioning category. | ||
Key terms
- Agentic Commerce: Agentic commerce is a buying and transaction model where software agents act on behalf of a person. The identity challenge is not just proving who owns the account, but constraining what the agent may do, for how long, and under what revocation and audit rules.
- Signal Blackout: A loss of the normal evidence used to make trust decisions, such as device fingerprints, cookies, and IP reputation. In agentic commerce, signal blackout creates ambiguity for fraud systems because automation can be both legitimate and hard to distinguish from abuse.
- Delegated Authority Model: A delegated authority model defines who is allowed to approve, review, or execute control-related decisions across the enterprise. It helps ensure requests reach the correct responsible party, especially when control owners, managers, and process owners sit in different teams, regions, or systems.
- Fraud Provenance: Fraud provenance is the record of how a transaction was formed, who or what initiated it, and which signals supported the decision to approve or decline it. Strong provenance allows investigators to reconstruct agent-led purchases and distinguish legitimate delegation from abusive automation.
Deepen your knowledge
The NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, machine identity security and secrets management. It helps practitioners align identity controls with the broader access decisions their programmes now face.
Published by the NHIMG editorial team on July 10, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org