The European Commission published its Business Wallet proposal in November 2025. It is a serious piece of infrastructure work. Cryptographic attestations, fine-grained access control, cross-border legal identity. Good stuff. But there is a meaningful gap in it, and it is one that is going to matter very quickly.
The regulation was written for humans. Specifically, for a human employee sitting in front of a screen, authenticating into a wallet, presenting credentials, and signing documents. That is a reasonable assumption for 2022. In 2026, it is already incomplete.
Agents are not a future thing. They are filing, signing, submitting, and transacting on behalf of businesses right now. And the Business Wallet regulation, as written, has no formal concept of what happens when an agent, not a human, knocks on the wallet's door.
We think that needs to change. So we submitted four formal comments to the Commission. Here is what we said.
Our comments to the Commission
Delegation chains need to be first-class citizens
Section 12 lays out a solid access control framework. Roles, mandates, powers of attorney. But it assumes a human is the one presenting those credentials. When an AI agent does it instead, there is no defined mechanism for verifying that the agent was actually authorized to act, what constraints it is operating under, or when that authority expires. We are asking the Commission to define “delegated agent access” as a distinct authorization mode, with machine-readable delegation policies bound to the wallet unit and automatic revocation when those policies lapse.
Transaction logs should capture the whole picture
The logging requirements in Section 7 are thorough for human-initiated transactions. They log who did what, when, with whom. That is exactly what is needed. What is missing is the agent layer: who authorized the agent, under what delegation policy, and across what chain of agents if there were multiple. Multi-agent workflows are increasingly common. Without logging the chain of custody, audit trails become incomplete and legal proceedings become complicated.
Agents need verifiable identity when communicating legally
The secure legal communication channel is a strong feature. But when an agent transmits a legal document on behalf of a wallet holder, who is it presenting itself as? The regulation does not answer that. We are asking for a standard for verifiable agent identity credentials, linked back to the wallet unit of the principal, and technically harmonized with MCP, the protocol that has become the de facto connectivity standard between AI agents and external systems.
The regulation needs a new actor: the Trusted AI Agent
The current actor model has three roles: wallet provider, wallet owner, and relying party. That is it. An AI agent acting on behalf of a business has no defined status. We are proposing a fourth: the Trusted AI Agent, defined as an autonomous system with a cryptographically bound, auditable delegation from the wallet owner. Without this, the Business Wallet cannot be the foundational trust layer for the Agentic economy, which would be a missed opportunity, given how close the existing infrastructure already is to making that possible.
Why this matters beyond compliance
Every enterprise deploying AI agents is currently solving the delegation and trust problem on their own, in isolation, inconsistently. That is exactly the pattern that produces security incidents, audit failures, and legal disputes.
The EU has a chance to standardize this at the infrastructure level before fragmentation sets in. The Business Wallet is the right vehicle. The eIDAS trust framework is already there. The cryptographic primitives are solid. What is missing is the agent layer, and that is exactly the gap nulla is built to fill.
These are not hypothetical concerns. The regulatory window for input is now. Once the implementing acts are drafted, the architecture gets locked in.