Adopt AI without losing control.

No. The goal is to make AI use safe and approved, so teams can use it with confidence instead of out of sight. Most organisations already use AI in more places than they realise. Citadel brings that use into the open and puts sensible limits in place.

Yes. When data must not leave your organisation, we design and deploy private AI on open-weight models in your own cloud tenant or on premises. This works well for internal search across documents, assistants for engineers and analysis of sensitive data.

It depends on how you use AI. Most everyday use, such as writing assistance, carries light obligations, while AI used in areas like recruitment, credit decisions or safety components faces strict requirements. The AI Risk Review classifies your use cases and shows which obligations apply.

Guardrails sit between your people or agents and the AI models. They detect and remove sensitive data from prompts, block requests that break policy, limit what AI agents are allowed to do and log activity for review. They are set up to match the policies you agreed on, not a generic template.

Prompt injection is when hidden instructions in a document, email or web page trick an AI system into doing something it should not, such as revealing data or taking an action. It becomes a real risk when AI agents can read outside content and act on your systems. Citadel reviews those connections and adds controls that limit the damage.

Usually a few weeks, depending on how widely AI is used across your organisation. You receive a map of AI use and data flows, a risk assessment and a prioritised set of recommendations you can act on straight away.