I tried to forge my own AI agent's audit log

Description and use

<p>Half the organisations running AI agents have already had one step outside its<br> permissions. Nearly half of the agents in production have no monitoring at all.<br> Only about a fifth treat an agent as something with an identity of its own.<br> Those are numbers from the Cloud Security Alliance and the 2026 State of AI<br> Agent Security report, and they describe the same gap from three directions:<br> agents act, and nobody can say afterwards what exactly they did.</p> <p>I spent a few weeks building the missing piece, and then spent an afternoon<br> attacking it. This is what survived.<

Problem

<p>Half the organisations running AI agents have already had one step outside its<br> permissions. Nearly half of the agents in production have no monitoring at all.<br> Only about a fifth treat an agent as something with an identity of its own.<br> Those are numbers from the Cloud Security Alliance and the 2026 State of AI<br> Agent Security report, and they describe the same gap from three directions:<br> agents act, and nobody can say afterwards what exactly they did.</p> <p>I spent a few weeks building the missing piece, and then spent an afternoon<br> attacking it. This is what survived.<

Outcome

<p>Half the organisations running AI agents have already had one step outside its<br> permissions. Nearly half of the agents in production have no monitoring at all.<br> Only about a fifth treat an agent as something with an identity of its own.<br> Those are numbers from the Cloud Security Alliance and the 2026 State of AI<br> Agent Security report, and they describe the same gap from three directions:<br> agents act, and nobody can say afterwards what exactly they did.</p> <p>I spent a few weeks building the missing piece, and then spent an afternoon<br> attacking it. This is what survived.<

Evidence

I tried to forge my own AI agent's audit log

Source