AI risk assessment
4–6 weeksSystem inventory, risk tiering, impact assessment and a prioritised remediation roadmap signed off by legal, security and the business owner.
- Model & data inventory
- Fundamental rights impact
- Risk tier classification
Services
Four engagements that cover the AI lifecycle.
Each engagement stands alone and composes into a full assurance programme. Fixed scope, fixed fee, evidence you own.
Independent model audit
6–10 weeksThird-party evaluation of performance, bias, robustness and explainability, with an evidence pack written for auditors and regulators.
- Bias & fairness testing
- Robustness and drift
- Documentation review
Red-teaming & evaluations
3–5 weeksStructured adversarial testing of LLM and agentic systems: prompt injection, data exfiltration, tool misuse and unsafe output paths.
- Jailbreak & injection suites
- Agent tool-abuse testing
- Eval harness handover
Governance operating model
OngoingThe committees, gates, policies and metrics that make oversight repeatable — plus training so your teams run it without us.
- Policy & control library
- Approval gates
- Board-level reporting