Deadline: 24-Mar-2026
The Advanced Research and Invention Agency (ARIA) is funding ambitious R&D projects to strengthen secure agent-to-agent interactions in untrusted environments. Through its Scaling Trust Programme, ARIA supports open-source tooling and fundamental research that enables provable AI security, protocol verification, and cyber-physical trust primitives. Applications close on 24 March 2026 at 14:00 GMT.
Programme Overview
The Advanced Research and Invention Agency (ARIA) is accepting applications under its Scaling Trust Research and Development Grant Programme. The initiative aims to build secure, trustworthy multi-agent systems capable of operating in adversarial or untrusted digital and cyber-physical environments.
The programme focuses on two main pillars:
• Open-source tooling for agent coordination and security
• Foundational research that converts empirical AI security into provable guarantees
The objective is to deliver real-world demonstrations, measurable trustworthiness, and high-impact adoption across academia and industry.
What Problem Is ARIA Addressing?
As AI agents increasingly negotiate, coordinate, and act autonomously, secure agent-to-agent interaction becomes critical. In untrusted environments, agents must:
• Translate vague human intent into formal security requirements
• Negotiate collective policies
• Generate and verify secure protocols
• Provide auditable execution records
ARIA’s Scaling Trust Programme aims to establish theoretical and practical foundations that ensure these processes are verifiable, secure, and interoperable.
Programme Structure: Three Interlinked Tracks
Although the programme includes three tracks, this funding call targets Track 2 and Track 3.
Track 1: Arena (Testing Environment)
A live adversarial environment where tools and research outputs are stress-tested.
Track 2: Tooling (Open-Source Infrastructure)
Supports development of reusable open-source components for secure agent coordination.
Track 3: Fundamental Research
Develops theoretical foundations and new security primitives for AI agent systems.
Track 2: Tooling (Open-Source Agent Infrastructure)
Track 2 funds the development of open-source agents and reusable components that act as baseline participants in the Arena.
Eligible Technical Areas
• Requirement gathering tools that convert ambiguous user input into formal security policies
• Negotiation engines that derive shared collective policies
• Security reasoning systems that generate and implement secure protocols
• Reporting and auditing tools that convert execution traces into concise compliance statements
Expected Project Outcomes
Projects must demonstrate:
• Competitiveness within the Arena
• Generality across multiple tasks
• Computational efficiency
• Evidence of community adoption
Funding and Duration
• £200,000 to £2 million per project
• Project duration: 3 to 12 months
• 4 to 6 teams expected to be funded
Track 3: Fundamental Research
Track 3 advances scientific confidence in AI agent security and coordination through theoretical and experimental research.
Core Research Areas
• Formal AI Security with provable guarantees
• Cyber-Physical Trust Primitives using physical or biological trust anchors
• Foundations of Generative Security for automated protocol creation and verification
• Bluesky research exploring novel or unforeseen theoretical challenges
Funding and Duration
• £100,000 to £3 million per project
• Project duration: 6 to 18 months
• 3 large research centres plus smaller exploratory teams
All outputs must be openly published to seed research communities and inform future Arena challenges.
Open-Source and Ecosystem Requirements
All funded software and research outputs must:
• Be released under permissive open-source licences
• Ensure transparency and interoperability
• Support long-term ecosystem growth
The programme operates through iterative phases including bootstrap, testing, improvement, and scaling, with quarterly milestones, check-ins, build weeks, and hackathons.
Who Is Eligible?
ARIA encourages applications from:
• Research institutions
• Universities
• Startups and technology companies
• Open-source developers
• Interdisciplinary AI security teams
Non-UK teams may be funded if they significantly enhance UK benefit and ecosystem impact.
How to Apply
-
Identify the appropriate track (Track 2: Tooling or Track 3: Fundamental Research).
-
Define the technical contribution and its transformative potential.
-
Demonstrate differentiation from existing approaches.
-
Provide a clear roadmap with milestones and measurable outputs.
-
Outline responsible innovation practices and risk mitigation.
-
Highlight UK benefit and ecosystem contribution.
-
Submit the proposal before the deadline.
Key Deadline
Applications close on 24 March 2026 at 14:00 GMT.
Awards are expected to be confirmed by June 2026.
Evaluation Criteria
Proposals are assessed based on:
• Transformative potential
• Technical differentiation
• Clarity and feasibility
• Responsible and secure approach
• Team motivation and expertise
• Benefit to the UK innovation ecosystem
Common Mistakes to Avoid
• Failing to demonstrate open-source commitment
• Lack of measurable milestones
• Weak differentiation from existing research
• Insufficient explanation of security guarantees
• Ignoring UK ecosystem impact
Why This Programme Matters
Secure multi-agent systems are foundational to future AI infrastructure, autonomous coordination, cyber-physical systems, and digital governance. By funding both practical tooling and theoretical breakthroughs, ARIA aims to create globally leading, provably secure agent ecosystems rooted in open research and real-world validation.
FAQs
1. What is the primary goal of the Scaling Trust Programme?
To strengthen secure agent-to-agent interactions in untrusted environments through open-source tooling and foundational research.
2. Which tracks are open in this call?
Track 2 (Tooling) and Track 3 (Fundamental Research).
3. What is the maximum funding available?
Up to £2 million for Track 2 and up to £3 million for Track 3.
4. Are international applicants eligible?
Yes, exceptional non-UK teams may be funded if they significantly benefit the UK.
5. Are outputs required to be open source?
Yes. All funded outputs must be released under permissive open-source licences.
6. What is the project duration?
3–12 months for Track 2 and 6–18 months for Track 3.
7. When is the deadline?
24 March 2026 at 14:00 GMT.
Conclusion
The ARIA Scaling Trust R&D Grant Programme represents a major investment in provable AI security, open-source coordination infrastructure, and next-generation trust primitives. By combining tooling, theoretical research, and adversarial testing, the programme aims to build trustworthy multi-agent ecosystems with global impact. Applicants with ambitious, high-impact ideas in AI security and coordination should prepare proposals before the March 2026 deadline.
For more information, visit ARIA.
