AI agent hardening and control

Reduce unwanted actions and excessive permissions in an AI agent, and test what happens when something goes wrong.

From €12,500 excl. VAT · 3–5 weeks

A good fit when

Teams with a concrete agent, an accessible repository and defined operational responsibilities.

The agent works in ordinary tests, but malicious inputs, permissions, failures and recovery have not been checked adequately.

The findings, with any vulnerabilities ranked by risk, and the agreed fixes to address them.

The classes that apply to your agent, within the agreed scope:

Hidden instructions in emails, documents or pages the agent reads (direct and indirect prompt injection)
Permissions and tools beyond what the task needs: unauthorised actions, access to other customers’ data or accounts
Misuse of connected tools, including through MCP (the protocol that connects the agent to tools), and chains of actions that bypass controls
Confidential data leaking into answers, logs or external systems
High-impact actions, such as payments, refunds or data changes, without human confirmation
Reconstructing failures: logs, traceability and recovery

Fifteen years building software people rely on.

These results describe my experience. For your project, we define goals and measures around your company’s process.

Assessment and agreed fixes · Remote · one application or agent

50% at start · 30% on delivery of findings · 20% on retesting

Commercial tools, environments and travel excluded

One retest of agreed fixes within 30 days

Initial technical assessment within 2 business days

Limit what an AI agent can do, test its defences and reconstruct failures before trusting it with real data and operations.

B2B service. A request creates no commitment: price and scope become binding only in an accepted written proposal.