Originally published by ARIA on 11 July 2025. Since 3 August 2026, Trust Everything, Everywhere is a broader opportunity space shared with the Safeguarded AI programme (see below). Read the original v1.0 (DOI record) or the current opportunity space.

Version 1 of our opportunity space, Trust Everything, Everywhere, is now live!

We’re focusing on new security primitives that straddle digital and physical worlds, and will be spending the next months zeroing in on a £10-100m first challenge within the space.

Please reach out and let us know what you think! Criticisms and questions especially welcome: we need perspectives from various disciplines and vantage points to harden our focus and help us steer in impactful directions.

Summary

Software is transcending digital screens to shape the physical world, connecting cyber instructions to real-world outcomes. We are entering a new cyber-physical age that can bring great benefits for humanity: interactions like bioprinting medicines from an online file, tasking an embodied AI agent to run errands for us, or outsourcing an experiment to a self-driving lab are nearly within reach.

Yet, without the right trust infrastructure, we risk every interaction to incur a steep ’trust overhead’, making it slower, costlier, and less certain or worse, not happen at all.

The solution is a new trust infrastructure that extends formal security reasoning into the physical world. This will unlock trillion-pound cyber-physical markets and a society free to trust everything, everywhere.

Beliefs

1. Creating a new trust infrastructure will unlock the cyber-physical economy. The trust building blocks that enabled today’s £24-trillion digital economy, like encryption and cryptographic signatures, do not extend to the physical world.
2. Formal security reasoning is key to designing new cyber-physical trust building blocks: rooted in different assumptions (e.g. quantum physics instead of computational hardness), and living in alternative informational substrates (e.g. DNA instead of bits).
3. AI will make secure cyber-physical interactions accessible to everyone. The next generation of trust infrastructure will allow AI systems to generate customised, on-demand protocols in situ for any given cyber-physical interaction, spanning atoms, molecules, waves, and bits.

Thanks to Sarath Murugan, Harry Jenkins, Ilan Gur, Edith-Clare Hall and Joshua Elliott for their help shaping this document, and to the many individuals both within ARIA and externally who were generous with their time and shared valuable feedback with us along the way. Onwards!

Update, 3 August 2026: a shared opportunity space

ARIA has revamped its opportunity spaces, the meta-structures that house programmes. Trust Everything, Everywhere got a makeover and now includes Nora Ammann & co’s Safeguarded AI programme. We are proud of this new version and its beliefs.

Three years in, ARIA’s Programme Directors kept converging on the same enabling technologies and scientific questions, so the original spaces have been regrouped into seven, each defined by the end state it could unlock rather than by discipline. The new Trust Everything, Everywhere brings together three of the original spaces:

• Trust Everything, Everywhere (v1, July 2025): this one, from which the Scaling Trust programme thesis followed in November 2025.
• Mathematics for Safe AI (David ‘davidad’ Dalrymple, November 2023): the original home of Safeguarded AI.
• Collective Flourishing (Nicole Wheeler, November 2025).

Scaling Trust and Safeguarded AI now sit side by side in the same space. The original beliefs still hold, and the v1 remit carries over in full to future programmes and seed funding. No active funding was affected.

The beliefs of the new space are broader. The cyber-physical thread above lives on in the third, alongside trust as the bottleneck of an AI-abundant world, a plurality of minds, and better trade-offs:

1. As AI makes creation abundant, trust becomes the bottleneck. Enormous effort is going into making things faster and cheaper — code, content, decisions, actions, connections. Far less goes into making those things trustworthy → justified trust, rather than raw capability, is where leverage lies.
2. The defining challenge of our time is how radically diverse minds live together. Humans differ in beliefs and values, and AIs differ from us and each other. The world must hold more views than ever → we seek to build the structures and tools that let a plurality of minds coexist and thrive, and resist the slide toward monoculture, centralised control or conflict.
3. Trustworthiness has to be engineered across the whole stack. As software moves beyond screens and into the world → trust infrastructure must extend into the physical, biological and institutional systems it acts through: chips, sensors, laboratories, neural interfaces, manufacturing processes, decision-making protocols, markets and norms. Everywhere minds and machines reach.
4. A new trust infrastructure can unlock better choices. Many of today’s hard trade-offs – security vs. freedom, speed vs. accountability, openness vs. control – are caused by constraints in our technological and institutional stack → the right building blocks can solve these tensions and increase our options.