UK's ARIA Opens £50M Fund for Trustworthy AI Agent Research

The UK's Advanced Research and Invention Agency has opened applications for its Scaling Trust programme, deploying nearly £50 million to solve one of AI's most pressing unsolved problems: how do you make autonomous AI agents trust each other?

Proposals are due March 24, 2026. Individual projects can request between £100,000 and £3 million for work spanning 3 to 18 months.

The Problem ARIA Wants Solved

As AI agents increasingly negotiate, transact, and coordinate on behalf of humans — booking flights, managing supply chains, executing financial trades — the infrastructure for secure multi-agent interaction barely exists. ARIA's programme targets the gap between today's empirical security guarantees and the provable, formal guarantees that high-stakes agent deployment requires.

The programme funds two tracks. Track 2 focuses on tooling: open-source agents and reusable components for secure requirement capture, negotiation, protocol generation, and verification in multi-agent settings. Track 3 targets fundamental research — formal AI security, generative protocol design and verification, and cyber-physical trust anchors.

Who ARIA Is Looking For

ARIA explicitly welcomes interdisciplinary teams bridging cryptography, AI security engineering, game theory, and robotics. Universities, research institutes, startups, established companies, and solo researchers are all eligible. While UK-based applicants are preferred — a strong benefit applies when 50% of project costs are spent in the UK — international applicants and mixed teams can apply.

Outputs should ideally be open-source, and a second solicitation is planned for mid-2027, creating a potential path for follow-on funding.

Why This Matters Beyond the UK

ARIA operates like a British DARPA — high-risk, high-reward research funding with minimal bureaucracy. For AI safety researchers, cryptographers, and multi-agent systems developers worldwide, this represents one of the largest dedicated funding pools for AI trust infrastructure.

The March 24 deadline is tight. Teams with existing work in verifiable computation, secure multi-party coordination, or formal methods for AI systems should review the full solicitation immediately. Tracking open international AI funding opportunities through platforms like Granted can surface similar programmes as they launch.