Expertise
Agentic AI safety
Governance for autonomous AI before it ships — loss-of-control, decision boundaries and human oversight.
The problem
Where the risk really sits
Agentic AI is being deployed into clinical decision support, drug discovery and other high-stakes workflows faster than governance matures. The risks are loss of control, unclear decision authority, and weak human-oversight mechanisms.
What DeTrust does
DeTrust conducts threat modelling and safety analysis of agentic AI in medical and pharmaceutical settings — examining loss-of-control risks, decision-authority boundaries and mechanisms for human oversight in collaborative AI environments.
The differentiator
Advice built by someone who engineered the methods
- Active threat-modelling and safety-analysis projects on agentic AI in the medical and pharmaceutical sectors (Honorary Research Fellow, WMG).
- Supervision of MSc research on risk-based agentic-AI governance and human–AI assurance.
- Framing relevant to NHS-adjacent clinical AI and MHRA-regulated AI medical-device assurance.
Selected research
From the research programme
Illustrations from the research programme
Intelligence-driven. Evidence first.
Tell the founder what you're working on. Every enquiry is read and answered personally.