How a tool or an API can tell an approved agent from a copied key, and notice when the agent has changed.
A company deploys agents that read tickets, query internal systems and act through tools. Each agent holds API keys. Security approved the agent in spring.
Since spring the system prompt changed four times, two tools were added, the model was upgraded, and one tool server changed its own description. The API key is the same as in spring. Every party still trusts the key, which proves that someone has the key.
The agent runs where its code, model reference, prompt and tool list can be measured. It receives a short-lived identity bound to those measurements. Tools check the identity and its claims on each call. When anything measured changes, the old identity lapses and the new one says what is different, which is the moment for a human or a policy to approve again.
An audit trail that names the agent version behind every action, and tools that refuse a copied credential.
An illustrative case built from public documentation and common architectures. It describes no particular company.
Working on attested ai agents? A first call takes 30 minutes and commits you to nothing. Bring your product and the one question that worries you.
Next fixed date: EU Digital Identity Wallets are due in every member state on 24 December 2026, 42 days from now.