The Race — Building Democracy in the Age of AI. People race toward a shared finish line.
The Race
← Trends

Established trend · Active

Vendors are defining the trust rails for agents and commerce

Latest assessment: · Evidence updated

Standing score
Unscored
Direction of power
Mixed-contested
Window
Unclear
Confidence
Moderate

What we think is happening

Identity, delegated permission and payment protocols are becoming recurring coordination layers through which agents gain the right to act. FIDO, Blueprint, UCP, Mastercard and Arcade show incumbents and specialist firms helping define this architecture.

Scope: Selected identity, permission and commerce architectures; interfaces are not market neutrality.

Potential direction of power

Common interfaces can enable action across providers, while admission, identity and transactions remain under the control of a few institutions. Independent governance, user revocation, cross-vendor choice and contestable pricing are not yet established consistently.

Supporting cases

Microsoft adds local AI execution controls while expanding Windows' AI role

Microsoft announced October 7 that Windows 11's Microsoft Execution Containment is generally available. It gives AI agents defined limits on access to files, networks and other resources, with agent identities and permissions. Several agent products can use it; some named integrations are still forthcoming.

The same announcement describes local/cloud combinations and forthcoming hardware. Experimental Copilot “Hydra Fusion” is a later-October preview, not an already available general service. The concrete current development is the execution-control layer; the broader promise of more local capability remains partly prospective.

For The Race, local execution can broaden options, while the operating-system owner acquires greater power to define trusted access and permissions. These two mechanisms point in different directions.

Distinct connection: Windows retains vendor-defined trust and permission controls. Overlapping evidence is not independent confirmation.

Read the case and sources →

Microsoft makes agent permissions enforceable outside the agent itself

Microsoft announced general availability of Microsoft Execution Containers on October 7. Developers and administrators specify which files, network connections and other resources a workload can use. The container enforces those limits outside the agent’s own control, so generated code cannot simply grant itself more access.

For example, a coding agent can be allowed to edit its repository while being blocked from personal documents or network destinations. The common configuration maps to different containment backends on Windows, macOS and Linux. Some options are platform-specific, and the hardware-isolated microVM option is experimental. Announced Entra identity, Intune management and expanded Agent 365 controls include future features; their availability is not established by the October announcement. Microsoft technical account ↗.

This gives the operator a more concrete way to limit delegated authority, while embedding an important control layer in Microsoft’s ecosystem. It is useful capability rather than proof that a particular organization has implemented effective oversight.

Read the case and sources →

European telecom regulators open a consultation on AI agents and internet openness

BEREC, the body coordinating European national telecom regulators, adopted a draft report at its October 1–2 meeting and discussed it publicly on October 7. It asks how generative AI changes people's ability to find information, reach services and choose providers. A formal public consultation runs through November 11; non-confidential contributions and a response summary are to be published. Official consultation ↗.

The practical concern is that an assistant may increasingly choose which services a person encounters, while a small number of companies control the models, cloud infrastructure and distribution channels behind it. The report considers whether existing European rules protect openness in those circumstances. It creates an opportunity to influence the analysis; it does not adopt new binding access or interoperability obligations. Primary draft ↗.

Read the case and sources →

A draft protocol specifies how personal agents could act with limited user permission

Personal Agent Protocol's public documentation now identifies draft version 0.1, updated October 9. Sierra announced the initiative on October 6 with Meta and other participating businesses. The draft describes how a user's AI assistant could identify itself to a business and ask to perform permitted tasks.

A business publishes a standard discovery file explaining its interfaces. A person can then authorize an agent with limited permissions through an authorization process, rather than handing over a full account password. The specification supports interaction through websites, APIs or business agents, with a route to human help. These are draft specifications: broad deployment, reliable cross-provider operation and actual purchasing permissions have not been demonstrated by their publication. Primary specification ↗; announcement and participants ↗.

The potential power benefit is that independent agents could reach businesses through a common interface. The unresolved question is who controls admission, identity and permitted actions. A common protocol can ease access while also creating another gatekeeper.

Read the case and sources →

Window of contestability

Unclear for users, developers and alternative providers: shared trust conventions can enable interoperability while their governance and admission rules can create new gates.

What would change this assessment

Open admission, independent participation, usable revocation and export, clear liability and successful provider switching would show whether common interfaces reduce gatekeeping.

Related cases

← All 24 trends