In recent months a quiet but significant phenomenon has been observed: more and more replies to tickets, emails and even commits on development platforms are generated by AI agents communicating with other AI agents, often without anyone explicitly disclosing it.
This is not a theoretical problem. If an AI agent replies to a customer, reviews a contract or interacts with a supplier without this being made transparent, who takes responsibility in case of error? And how do you know, as the recipient, whether you are actually talking to a person?
The European AI Act addresses exactly this point (Art. 50): anyone using AI systems that interact with people must disclose it, except in obvious cases. But the rule assumes that human-machine interaction is still the norm. When the agents themselves communicate with each other, without a human in the middle at every step, the question of transparency becomes more complicated.
For SMEs starting to introduce AI-based automation tools, the practical question is this: do you know exactly in which processes your company already generates fully automated communications today, and whether this is disclosed, documented and traceable?
I use AI intensively in my own work, but with a non-negotiable principle: every source, every draft, every final decision passes through my direct control. This is exactly where AI governance is really decided.
Related: AI Governance Checklist (free · 10 questions · 5 minutes)
Research as of: September 1, 2026 — Regulation evolves quickly; this post does not replace individual legal or professional advice.
