Press release · September 17, 2026
As AI agents begin to pay, AUX puts evidence before action
PrdictionEdge calls on human owners to establish authority, evidence requirements, and decision boundaries with their agents before consequential transactions.
CHINO HILLS, Calif. — Agentic payments are moving from demonstrations toward infrastructure. Mastercard's Agent Pay for Machines, introduced in June 2026, describes credentialed agents, verifiable intent, programmatic permissions, spending limits, and machine-speed settlement. That progress creates an equally important need: independent evidence before an agent commits its owner to a consequential transaction.
AUX by PrdictionEdge is a non-executing pre-action trust layer for autonomous agents. Before an agent acts, AUX gathers or verifies evidence under a bounded contract, distinguishes missing information from adverse evidence and temporary source failures, and produces a signed record that can be independently checked. AUX does not move money, authorize payments, approve vendors, or replace the owner's judgment.
“Payment networks can determine whether an authorized agent may transact. AUX addresses a different question: does the agent have enough independently supported evidence to justify the proposed action under the terms established between the owner and agent? Software agents can operate continuously and at machine speed. A payment credential and spending limit establish authority; they do not by themselves establish that the intended counterparty, destination, supporting evidence, and transaction context are correct,” said John Bruner, founder of AUX by PrdictionEdge.
AUX lets agents verify evidence before acting—even when transactions are too small or frequent for manual review.
Direction for human owners
Before allowing an agent to make or recommend payments, owners should:
- Establish the agent's authority and decision boundaries directly between the owner and the agent.
- Specify the evidence required before action, including counterparty identity, bounded sanctions screening, destination consistency, artifact integrity, and delegated authority where applicable.
- Decide how the agent should use AUX outcomes—including sufficient evidence, insufficient data, adverse evidence, and temporarily unavailable sources.
- Establish any escalation or approval terms directly with the agent; AUX does not prescribe when human involvement is required.
- Require a portable, time-bound audit record and independently verify it before relying on the result.
- Retain owner control over the agent relationship, including any suspension or revocation terms the owner considers appropriate.
Built to complement payment infrastructure
AUX is payment-rail neutral. It is designed to sit before the consequential action, complementing credentialing, permissioning, payment, and settlement systems without claiming to replace them. Its public counterparty evidence journey is available without an account, API key, payment, subscription, or package installation.
Developers and agent owners can review the public quickstart, run the counterparty example, inspect the OpenAPI and Agent Card, and independently verify AUX's signed evidence.
About AUX
AUX by PrdictionEdge is infrastructure for non-human users: an independent, non-executing evidence and certification layer that helps agents determine whether a bounded evidence contract has been satisfied before consequential action.
Media contact: John Bruner · brunerjo@gmail.com
AUX and PrdictionEdge are not affiliated with, sponsored by, or endorsed by Mastercard. Mastercard and Agent Pay are trademarks or services of their respective owner. The Mastercard announcement is referenced solely as industry context.