ARIA Safeguarded AI Programme

ARIA’s Safeguarded AI Programme sits under the Advanced Research and Invention Agency and focuses on mathematical guarantees for AI safety. It is a UK programme led by David Dalrymple, structured as a convertible award rather than a plain grant. The funding range runs from £200,000 to £5 million, with a median actual award of £1.2 million. The program is open to for-profit R&D performers in the UK and internationally. Nonprofits and individuals are excluded, the TRL window is 2 to 5, and the same solicitation cannot be stacked with another award. The annual cycle and convertible structure make it closer to a mission-driven investment route than a standard research grant. Applicants that fit best are teams working on hard safety properties, proofs, or verification methods that can be tied to practical AI systems. The programme’s emphasis is not on generic model building but on demonstrable safeguards that can be articulated and tested. That makes it especially relevant for groups that can bridge theory, software, and implementation within a single proposal.

Each grant below is a distinct funding opportunity with its own eligibility, scope, and deliverables.