Home / Our roles / AI Security & Compliance
Security & ComplianceWhat the system is not allowed to do
An organisation answers for the software it puts online to speak in its name. This trade turns that responsibility into technical constraints you can verify.
Why this role
In February 2024, a court ordered an airline to honour a discount invented by its conversational agent. The compliance of an AI system is not settled in a contractual appendix but in what the code allows.
What it does concretely
- Write down what the system is not allowed to produce, and enforce it server side
- Isolate conversations and data outside the web root, with short retention
- Set quotas and logging before going live, not after the incident
- Check where the data lives, and be able to prove it
On a sales agent we operate, a single rate can be quoted: any other amount in euros triggers the replacement of the whole answer by the server, before it reaches the screen.
What it is not
This is not an audit that produces a report, nor a box to tick before going into production.
Two ways to access it
Inside your team
A profile joins your team for an agreed period, works on your tools and your data, and passes on what they build as they go.
On a deliverable
The role is mobilised on a project with a defined scope, alongside the other skills required. You receive a deliverable, not a report.
Does this role match your need?
We start with a free diagnostic: what can be automated, what must not be, and the profile that makes sense at your end.
Book a diagnostic