SERVICES
Ethical AI and Data Integrity
Governance that produces a defensible answer, not a policy nobody reads.
Short answer
Ethical AI in practice means knowing which AI systems your organisation operates, what data they use, who is accountable for each, and what happens when one is wrong. The EU AI Act (Regulation (EU) 2024/1689) places duties on providers and on organisations deploying AI professionally, which makes this an operational requirement rather than a values statement.
How it actually works
Almost every ethical failure we see with AI in a company is a governance gap. Nobody owned the system. Nobody reviewed its output. Nobody had written down what wrong looked like. The model was rarely the problem; the absence of anyone responsible for it was.
The EU AI Act turns that into an operational requirement rather than a values statement. Since February 2025, organisations deploying AI have had a literacy obligation. Since 2 August 2026, Article 50 requires chatbots to disclose themselves and AI-generated content to be labelled. From December 2027, high-risk systems — recruitment screening above all — carry documentation, oversight and risk-management duties. The penalties reach 7% of worldwide turnover.
What we do is unglamorous and specific. We inventory every AI system you actually run, including the features inside software you already pay for. We classify each one and name who is accountable. We fix the disclosures that are already late. And we produce the documentation your counsel will ask for, and tell you plainly where a lawyer needs to be involved instead of us.
The Digital Omnibus, approved 29 June 2026, deferred high-risk obligations to December 2027. It did not defer Article 50, which has applied since 2 August 2026.
What the engagement covers
- System inventory
- Every AI system in use, its purpose, its data, and a named accountable owner. Most organisations discover systems they did not know were running.
- Risk classification
- Where each system sits under the Act’s risk tiers, and which obligations follow.
- Data integrity
- What goes into a model, where it came from, and whether you had the right to use it.
- Failure handling
- Human review, escalation and a documented path for a wrong output. Most incidents are process failures, not model failures.
How it runs
- Inventory what is running and who owns it.
- Classify by risk and identify the obligations that follow.
- Close the documentation and review gaps.
- Set a review cadence, because the systems change.
Is it for you?
It is, when
- You have a chatbot or AI assistant in front of customers and are not certain it discloses itself.
- You use AI anywhere in recruitment or employee assessment.
- Nobody in the company could list every AI system in use if asked today.
It is not, when
- You need a legal opinion. We are engineers; we prepare the material and know when to stop.
Frequently asked
Does the EU AI Act apply to a small company?
It places duties on providers of AI systems and on organisations deploying them in a professional context, with obligations scaled to risk. A small company using a general assistant carries far less than one deploying AI in hiring or credit, but "far less" is not "none".
Is this a legal service?
No. We are engineers, not lawyers. We map what you run, classify it, and produce the documentation and controls your counsel will ask for. Legal interpretation stays with your counsel.
Where does the ethics part come in?
In the same place as the compliance part. Almost every ethical failure we see is a governance gap: nobody owned the system, nobody reviewed the output, and nobody had defined what wrong looked like.
Contact us
Send us the brief or book a short call. You get concrete next steps in reply.