---
title: "OpenAI Rolls Out ChatGPT Wallet Feature with Human Oversight Requirements"
url: https://newscentral.site/openai-rolls-out-chatgpt-wallet-feature-with-human-oversight/
language: en
publisher: "News Central Site"
section: "Business"
published: 2026-11-21T11:00:00.000Z
updated: 2026-11-21T13:49:51.121Z
id: 41cc2823-ad63-4972-a5f3-35347397734a
source: "Fortune https://fortune.com/2026/11/21/chatgpt-wallet-ai-agents-payments-liability-mandates/"
attribution: "Link to https://newscentral.site/openai-rolls-out-chatgpt-wallet-feature-with-human-oversight/ and name News Central Site when you quote or summarize this story."
---

# OpenAI Rolls Out ChatGPT Wallet Feature with Human Oversight Requirements

The feature allows AI agents to make payments on behalf of users, but requires human approval before finalization.

OpenAI has rolled out ChatGPT Wallet, a feature allowing its AI agents to make payments on behalf of users, but these transactions still require human oversight before being finalized.

While this safeguard may provide some initial security, it's ultimately a limitation that hinders the true potential of agentic payments. For an agent to be truly effective, it should be able to make decisions independently, without needing constant human approval at every stage.

The technology behind agentic payments is rapidly advancing, with significant progress made in integrating existing payment methods into AI systems. However, there remains a crucial question that has yet to be addressed: what are the limits of an agent's decision-making authority?

For instance, consider a scenario where an AI agent is tasked with planning a family vacation within a set budget. The agent books flights, hotels, and other arrangements without issue, but ultimately delivers a trip that falls short of the user's expectations.

The issue of accountability arises when AI agents fail to deliver as expected, leaving users wondering who is responsible for their decisions. This problem persists even when merchants and payment providers follow existing dispute procedures, which may offer some remedies but do not clearly assign blame.

Authorization to spend does not necessarily mean authorization to exercise judgment, a distinction that current infrastructure fails to recognize. The industry has begun to develop verifiable mandates to address this gap, with Google's Agent Payments Protocol being one notable example.

This protocol uses cryptographically signed mandates to record user requests and approvals, but it only addresses half the problem. A signature can prove what was authorized, but it cannot confirm whether the agent interpreted the user's preferences correctly or acted in their best interest.

Verifying an AI agent's actions is not a straightforward task. It requires three key layers of verification: mandates that capture intent, mechanisms to enforce those intentions, and safeguards against outside incentives.

Mandates that encode intent are essential for verifiable transactions. They should include the user's priorities, trade-offs, and conditions for escalation, such as requiring permission before making non-refundable purchases. These constraints must be verifiable across the transaction, involving multiple parties without compromising user privacy.

Ensuring AI agents act in accordance with their mandates is a complex challenge that demands innovative solutions. By developing more robust mechanisms to verify agent actions, the industry can build trust and confidence in these emerging technologies.

To resolve disputes arising from agent actions, a clear liability framework must be established. This framework would determine which party bears responsibility when an agent acts within its spending authority but outside its intended purpose. Currently, ambiguity in instructions or misinterpretation by the provider can lead to losses for various stakeholders, including users, providers, merchants, and payment processors.

The allocation of liability should be settled beforehand, ideally by identifying the party that can prevent harm at the lowest cost. This approach has been successfully implemented in other industries, such as the card industry's EMV liability shift, which incentivized merchants to adopt chip technology to reduce counterfeit-fraud losses.

A similar principle is applied in product liability law, where defects are allocated along the supply chain. By establishing clear allocation of risk, it becomes possible to price and insure risks that can be assigned. This market creation enables stakeholders to manage and mitigate potential losses.

In the context of AI-powered agents, a duty of loyalty must also be considered. Currently, OpenAI does not take a commission on Wallet transactions, but business models can evolve over time. As an agent's platform earns revenue from merchants through fees or advertising, it may face conflicts of interest that are invisible to users.

These conflicts can be mitigated by enforcing standards for consumer agents, including disclosure requirements and best-interest obligations. Independent verification mechanisms must also be in place to ensure compliance with the user's mandate.

The development of agentic protocols is underway in AI labs, payment companies, and commerce platforms. These entities are well-positioned to create private standards for identity verification, credentials, and payment flows due to the benefits of interoperability for all parties involved.

As someone who has worked on both sides of regulatory rulebooks, it's clear that settling liability in advance is crucial for systems to scale. In my experience as chair of the FATF’s Risk Group and a fintech executive, rules that clearly assign responsibility for failure have been the most effective.

Companies that establish clear liability rules, loyalty mechanisms, and mandates within their agents will be ahead of the curve. These entities will earn consumer trust by demonstrating accountability when something goes wrong.

The same principle applies to card payments, which became global infrastructure because consumers believed someone would take responsibility for any issues. Agentic payments require a similar promise of accountability across all stakeholders, consumers, AI platforms, merchants, and financial institutions.

In essence, whoever makes this promise credible will not only facilitate agentic payments but also play a key role in shaping the rules of the new economy.

The concept of agentic payments has sparked debate about accountability and responsibility in a world where AI agents make financial decisions on behalf of individuals.

As these AI agents handle transactions autonomously, questions arise regarding who should be held accountable for their choices, including any potential mistakes or unethical actions. In essence, whoever makes this promise credible will not only facilitate agentic payments but also play a key role in shaping the rules of the new economy.

The uncertainty surrounding accountability highlights the need for clear regulations and standards governing AI decision-making processes. This lack of clarity creates an environment where individuals may be left vulnerable to financial risks associated with AI-driven transactions, further underscoring the importance of establishing robust frameworks for agentic payments.

---
Source: [Fortune](https://fortune.com/2026/11/21/chatgpt-wallet-ai-agents-payments-liability-mandates/)  
Published by News Central Site: https://newscentral.site/openai-rolls-out-chatgpt-wallet-feature-with-human-oversight/
