GCRI · Foundations
Technical trust infrastructure for a more prepared world
Risk now crosses the ledgers that institutions keep. This essay sets out what technical trust infrastructure is, why the evidence shows it is needed, and how the programme builds it without claiming any authority it does not hold.
- As of
- 21 October 2026
- Reading
- 12 minutes
- Evidence
- 16 sources, cited in order
Methods and evidence
- Evidence
- 16 sources, cited in order and listed under References
- Figures
- 5 figures
- Data availability
- Every chart's data is published as a table on this page.
01What technical trust infrastructure is
What technical trust infrastructure is
Technical trust infrastructure is the layer of records, methods, interfaces and controls that sits between a technical capability and the decision it informs. It does not make a model accurate, a dataset complete or a dashboard right. It makes each of them legible: what was built, on which data, under which assumptions, with what uncertainty, by whom, under which agreement, and what changed when an error was found. A capability that carries that record can be compared, challenged and reused; one that does not can only be believed.
The Global Risks Forum (GRF) and The Global Risks Alliance (GRA) work with GCRI in the Nexus Consortium, and each reads the same record for a different purpose: GCRI for science and method, GRF for convening and the governance of Nexus Universe, GRA for whether finance and insurance can read the evidence 2,7. The record is the shared object. The decision is not shared: it stays with the government, regulator, operator or financial institution that holds the mandate, as the Decide step of the Nexus cycle states 2.
The earlier version of this page described the same ambition in the language of a programme then being designed. Several structures it named, a branded annual compute core, a passport for every component and separate continuity rails, are not on the current record, and this version does not use them. What is on record is simpler and stricter: twenty systems named by Nexus Standards display codes, nine evidence labels, seven programme states and service levels from R0 to R5 1.
02Why risk no longer stays in its ledger
Why risk no longer stays in its ledger
Institutions keep ledgers by mandate: a ministry counts public assets, an insurer counts insured claims, a utility counts outage minutes, a hospital counts admissions. Systemic risk is the part of loss that travels between those ledgers. The Global Assessment Report 2025 of the UN Office for Disaster Risk Reduction puts direct disaster costs at about $202 billion, and the true cost at more than $2.3 trillion once cascading and ecosystem costs are taken into account 3. The burden, and the debt it creates, falls disproportionately on developing countries 3.
The mechanism is well described in science. In coupled networks, the failure of a node in one network can disable dependent nodes in another, so that a small initial failure propagates in a way no single network would predict (Buldyrev et al., 2010) 8. Electricity, water, telecommunications, payments and logistics are coupled in exactly this way. The practical consequence is that no single holder of evidence can see the whole cascade, and every holder needs to trust evidence it did not produce.
That trust cannot rest on appearance. A persuasive dashboard, a fluent AI answer or a detailed simulation can each be wrong in ways its user cannot see. Science learned this about itself: a Nature survey of more than 1,500 researchers treated reproducibility as a problem the research system had to face openly (Baker, 2016) 9. Risk work that informs public decisions carries the same obligation, with higher stakes.
Fig. 01Most of the cost of a disaster is not in the first ledger
Figure 1 in words, and its data
Two bars on one US dollar scale. Direct disaster losses: about 202 billion dollars. Full cost including cascading and ecosystem effects: more than 2.3 trillion dollars a year. A side panel notes that the full cost is more than ten times the direct damage usually counted, and that the cost and the debt it creates fall disproportionately on developing countries.
03Capability is spreading faster than its reach
Capability is spreading faster than its reach
Early warning shows both the progress and the gap. The Global Status of Multi-Hazard Early Warning Systems 2025, published by the World Meteorological Organization with UNDRR, reports that 119 countries, 60% of all countries, now report a multi-hazard early warning system, a 113% increase in ten years 4. Only 43% of small island developing States do 4. Average comprehensiveness has risen 45% since 2015, and 72% in Africa, which nonetheless keeps the lowest scores 4. The report is the monitoring instrument for Target G of the Sendai Framework for Disaster Risk Reduction 2015–2030 and for Early Warnings for All 4,10.
Connectivity tells the same story. The International Telecommunication Union estimates that about 6 billion people use the Internet in 2025 and 2.2 billion do not 6. A capability that exists is not a capability that reaches people, and a capability that reaches people is not yet one whose outputs can be checked by those who must act on them. Technical trust infrastructure addresses the third gap, which is the least visible and the one that turns the other two into decisions.
Fig. 02Warning systems are spreading faster than their reach
Figure 2 in words, and its data
Left panel, share of countries reporting a multi-hazard early warning system in 2025: all countries 60 per cent (119 countries, a 113 per cent increase over ten years); small island developing States 43 per cent. Right panel, rise in the average comprehensiveness score since 2015: all regions 45 per cent; Africa 72 per cent, the largest gain, though Africa still has the lowest scores.
| Measure | Value (per cent) |
|---|---|
| All countries reporting | 60 4 |
| SIDS reporting | 43 4 |
| Comprehensiveness rise, all regions | 45 4 |
| Comprehensiveness rise, Africa | 72 4 |
04From a claim to a record that can be checked
From a claim to a record that can be checked
The programme answers the third gap with records, not reassurance. Every finding carries one of nine evidence labels: observed, published statistics, institution-reported, estimated, modelled, forecast, scenario, expert judgement or unknown 1. Missing data is shown as missing, never as zero, and simulated results are never reported as observed impact 1. The label "published statistics" names a source type; it is not an endorsement of the finding by the authority that published the data 1.
Each kind of technical claim then has a home in one of the twenty systems. Canonical data, provenance and risk sit in NXS-GRX (canonical data, provenance and risk); dashboards and workflows in NXS-STU (composition, dashboards and workflows); models and agents in NXS-NXQ (software, models and capability lifecycle); simulations and frontier evaluation in NXS-LAB (research, experiments and simulations); and every handoff to a receiving institution in NXS-RLS (participants, agreements and handoffs). NXS-NSF (trust, authorisation and assurance) spans all of them and decides who may see, change or release a record. These pages describe designs for reading; none records a deployed system 2.
Two further rules make the record durable. Every object of the programme year carries one identifier, a UNI under the proposed Nexus Standard NS-1, so it can be cited and versioned; and an accepted correction creates a new recorded version rather than a silent edit, as the identifiers and records page explains 1.
Fig. 03From a claim to a record that can be checked
Figure 3 in words
Six rows. Dashboard: data sources, update time, uncertainty and the label of each value, held in NXS-STU and NXS-GRX; a dashboard is a view, not a finding. AI model or agent: model role, version, data boundary, tool permissions and human review, in NXS-NXQ and NXS-NSF; an AI output is never evidence on its own. Simulation or twin: assumptions, inputs, exclusions, run conditions and method, in NXS-LAB and NXS-EOP (earth observation and digital twins); a simulated result is never reported as observed. Dataset: source, licence, provenance, location and boundary, in NXS-GRX; missing data is shown as missing. Exercise: scope, simulation marking, stop-work controls and after-action record, in NXS-DSS (decision cases and deliberation) and NXS-HUB (knowledge, records and learning); no exercise message is delivered as a warning. Handoff: what crossed, to whom and under which agreement and gate, in NXS-RLS; the receiving authority decides. NXS-NSF spans every row.
Fig. 04Nine labels that state the status of evidence
Figure 4 in words
A row of nine labelled boxes, numbered one to nine, under a bar shading from measured directly to not yet known: observed; published statistics; institution-reported; estimated; modelled; forecast; scenario; expert judgement; unknown. Below, three rules: every finding carries one label; missing data is shown as a gap, never as zero; modelled and scenario results never pass as observed impact.
05Trust is becoming a legal and technical requirement
Trust is becoming a legal and technical requirement
The demand for checkable records is no longer only a scientific norm. The FAIR principles asked in 2016 that research objects be findable, accessible, interoperable and reusable for machines and for people 11. The OECD AI Principles, adopted in 2019 and updated in May 2024, set an intergovernmental baseline for trustworthy AI 12. The NIST AI Risk Management Framework, released on 26 January 2023 and extended by a Generative AI Profile on 26 July 2024, gives organisations a voluntary method 13. Version 2.0 of the NIST Cybersecurity Framework, released on 26 February 2024, added emphasis on governance and supply chains 14.
Law is following. The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024; its prohibitions and AI literacy duties applied from 2 February 2025 and its general-purpose AI obligations from 2 August 2025 5. It became applicable on 2 August 2026, when its transparency rules and enforcement by the AI Office and national authorities began 5. After the simplification agreed in 2026, the high-risk rules for uses such as critical infrastructure, education and employment apply from 2 December 2027, and those for regulated products from 2 August 2028 5. High-risk systems will need logging for traceability, documentation, data quality and human oversight: the content of a technical record 5. Globally, the Global Digital Compact, negotiated by 193 Member States, commits governments to more investment in digital public goods and digital public infrastructure 15.
None of these instruments is applied by the programme on anyone's behalf. Each binds or guides its own adopters, and the programme's records are designed to be readable against them. The Nexus Standards for digital systems and AI are proposed, not adopted, and each jurisdiction decides whether to use them 2.
Fig. 05Trust is becoming a legal and technical requirement
Figure 5 in words
A horizontal timeline of ten dated instruments. March 2016: FAIR principles for findable, accessible, interoperable and reusable data. May 2019: OECD AI Principles, updated May 2024. January 2023: NIST AI Risk Management Framework 1.0. February 2024: NIST Cybersecurity Framework 2.0, adding governance. August 2024: EU AI Act, Regulation (EU) 2024/1689, enters into force. September 2024: Global Digital Compact, negotiated by 193 Member States. February 2025: AI Act prohibitions and AI literacy duties apply. August 2026: the AI Act becomes applicable, with transparency rules and enforcement. December 2027: high-risk rules for Annex III uses including critical infrastructure. August 2028: rules for high-risk systems in regulated products. A marker shows 21 October 2026.
06From innovation theatre to verifiable capability
From innovation theatre to verifiable capability
A pilot can be impressive and narrow, a demonstration persuasive and incomplete, a prototype promising and untested beyond one environment. The programme separates these states on purpose. In the Labs pillar, an experiment is exploratory and labelled as such; it becomes evidence only after the Foundry pillar has rebuilt it and a Nexus Universe mission has tested it 2. In the Foundry pillar, a configured or simulated capability is not a validated result 2.
Verifiable capability therefore does not mean that success is assured. It means a capability that can be described, tested, observed, recorded, challenged and corrected, with its maturity stated conservatively and its public claims bounded. The Labs gateway applies one protocol to every test: a use case and comparator, data and metrics, a reproducible build, a test envelope and findings that include failure cases. Testing is not certification and a testbed is not a regulatory sandbox or a procurement status 2.
07Contribution without capture
Contribution without capture
Much of the capability the programme needs belongs to others: computing centres, network operators, universities, data providers and technology firms. The question is not whether they take part but how they take part without capturing a public-good purpose. The answers are written into the model. Technology, infrastructure and service providers work with the Nexus Consortium at arm's length under written agreements and are described by role only 2. Evaluation never depends on a high contribution, offered capacity is never counted as reserved, consumed or funded, and sponsorship never influences a result 2.
The same rule protects public bodies. A regulator that takes part gains no approval and no privileged access; a city relationship never becomes national representation; an exercise message is never delivered as a warning 2. The claims and readiness register lists what each public statement may and may not say, and the conflicts and independence policy explains how interests are declared.
08What is changing
What is changing
- Cascading cost becomes visible. Counting cascading and ecosystem effects raises the estimated cost of disasters by an order of magnitude, which moves the question from response to shared evidence 3.
- Warning moves from systems to reach. With 119 countries reporting systems, the open problem is coverage where it is lowest, starting with small island developing States 4.
- AI records become legal records. From 2 December 2027, high-risk AI in critical infrastructure must carry logging, documentation and human oversight in the EU 5.
- Governance enters cybersecurity. CSF 2.0 makes governance an explicit part of cyber risk management, which ties technical controls to accountable owners 14.
- Digital public goods gain a political mandate. The Global Digital Compact gives governments a shared commitment to invest in digital public goods and infrastructure 15.
09Three institutions, one record, separate decisions
Three institutions, one record, separate decisions
The same record looks different from each institution, and all three readings are needed. GCRI reads it for science and method: its teams run the experiments of the Labs pillar, bring the methods and data the Foundry pillar builds on, and act as the technical office of Nexus Universe 2,16. GRF reads it for mandate and convening: it stewards the annual programme, records each gate from G0 to G7 and governs universe.nexus 7. GRA reads it for finance: it makes evidence readable to finance and insurance through finance readiness and states its regulated boundaries 2.
Each reading passes through the eight steps of the Nexus cycle: Question, Evidence, Build, Organize, Test, Publish, Decide and Renew. Seven steps run in the programme. Decide does not: it belongs to the competent authority, under its own rules 2. Risk Academy teaches the same discipline in the open, and Reports publishes it with sources, limits and corrections.
What this page does not claim: technical trust infrastructure, as described here, is a design and a discipline; no system, room or service is recorded as deployed, and the current service level is R0, information and enquiries. GCRI does not regulate or certify, GRF does not regulate, finance or procure, and GRA does not lend, underwrite, rate or invest. Naming an agency, a framework or a law is not a claim of partnership, endorsement or conformity. Every decision informed by programme evidence is taken by the competent authority, outside the programme; any finance vehicle is formed, if at all, by lawful actors outside it, as the lawful handoffs page explains.
Questions
Questions and answers
Is technical trust infrastructure a platform that GCRI sells or operates?
No. It is a set of records, methods and controls carried by the programme's systems, which are documented as designs. Ordinary participation has no fee, providers work at arm's length under written agreements, and no system is recorded as deployed or offered as a service.
Does a record in the programme prove that a capability works?
No. A record shows what was built, tested and observed, with its evidence label and limits. It allows others to check a claim; it does not certify the claim, and testing in the programme is never presented as certification or regulatory approval.
How does this relate to the EU AI Act or the NIST frameworks?
Those instruments bind or guide their own adopters. The programme designs its records so that they can be read against such instruments, but it applies none of them on anyone's behalf and makes no statement of conformity for any participant.
Who decides what happens with the evidence?
The competent authority, operator or licensed institution that holds the mandate, under its own rules. Decide is the one step of the Nexus cycle that sits outside the programme.
What can an institution do today?
At service level R0, it can read the architecture and current states and send a question through the enquiry form, which returns a reference and the next step. Applications, allocations and registrations open only when their team, terms and review capacity are in place.
Terms
Glossary
- Technical trust infrastructure
- The records, methods, interfaces and controls that make a technical capability legible, checkable and correctable before it informs a decision.
- Evidence label
- One of nine tags every finding carries, from observed to unknown, stating what kind of evidence it is.
- Cascading cost
- Loss that propagates from the first affected system to dependent systems, sectors or places, beyond the direct damage.
- Verifiable capability
- A capability that can be described, tested, observed, recorded, challenged and corrected, with its maturity stated conservatively.
- Service level R0
- The current state of the programme: information and enquiries, with a durable reference for each enquiry.
- UNI
- The identifier every programme object carries under the proposed Nexus Standard NS-1, so it can be cited, versioned and resolved.
Related
Further reading across the network
- The mission: verifiable technical infrastructure for an age of systemic risk
- The public-good technical stack
- The technical architecture of the twenty systems
- Universities, research labs and students
- STEM Commons
- Programme status
- Status and service levels in the documentation
- The participation records of the programme
Reuse and cite
Text CC BY 4.0. Cite the page with its date; data tables are linked under each figure.
Sources
References
16 sources, numbered in the order they are first cited
- The Global Centre for Risk and Innovation (2026-10). Programme status. therisk.global. Accessed 2026-10-21.
- The Global Centre for Risk and Innovation (2026-10). How Nexus Consortium works. therisk.global. Accessed 2026-10-21.
- United Nations Office for Disaster Risk Reduction (UNDRR) (2025-05-27). Global Assessment Report on Disaster Risk Reduction 2025: Resilience Pays, Financing and Investing for our Future (launch release). reliefweb.int. Accessed 2026-10-21.
- World Meteorological Organization (WMO) with UNDRR (2025-11-12). Early warning systems reach new heights, but critical gaps jeopardize global progress (Global Status of Multi-Hazard Early Warning Systems 2025). wmo.int. Accessed 2026-10-21.
- European Commission (2026-10). AI Act: regulatory framework for artificial intelligence. digital-strategy.ec.europa.eu. Accessed 2026-10-21.
- International Telecommunication Union (ITU) (2025-11-17). Facts and Figures 2025: global Internet use continues to rise but disparities remain. www.itu.int. Accessed 2026-10-21.
- The Global Risks Forum (2026-10). Nexus Universe. globalriskforum.com. Accessed 2026-10-21.
- Buldyrev et al., Nature 464, 1025 (2010-04-15). Catastrophic cascade of failures in interdependent networks. doi.org. Accessed 2026-10-21.
- Baker, Nature 533, 452 (2016-05-25). 1,500 scientists lift the lid on reproducibility. doi.org. Accessed 2026-10-21.
- United Nations General Assembly (2015-06-03). Sendai Framework for Disaster Risk Reduction 2015–2030 (A/RES/69/283). docs.un.org. Accessed 2026-10-21.
- Wilkinson et al., Scientific Data 3, 160018 (2016-03-15). The FAIR Guiding Principles for scientific data management and stewardship. doi.org. Accessed 2026-10-21.
- Organisation for Economic Co-operation and Development (OECD) (2024-05). OECD AI Principles. oecd.ai. Accessed 2026-10-21.
- U.S. National Institute of Standards and Technology (NIST) (2024-07-26). AI Risk Management Framework. www.nist.gov. Accessed 2026-10-21.
- U.S. National Institute of Standards and Technology (NIST) (2024-02-26). NIST releases version 2.0 of landmark Cybersecurity Framework. www.nist.gov. Accessed 2026-10-21.
- United Nations (2024-09). Global Digital Compact. www.un.org. Accessed 2026-10-21.
- The Global Centre for Risk and Innovation (2026-10). GCRI Canada. therisk.global. Accessed 2026-10-21.
Record
Editorial record
Facts and programme states are given as of 21 October 2026. This page is reviewed at least every 90 days and whenever a cited figure, instrument date or programme state changes. It replaces an earlier version of the same page, which described a branded annual compute core, component passports, continuity rails and an observatory product that are not on the programme's current record; those structures have been removed, and the AI Act dates have been updated to the timeline in force after July 2026.
Published by the GCRI editorial desk under the editorial integrity policy. To report an error or request a correction, use the corrections route; corrections are dated and listed with the page. Participation is recorded in the programme's participation records.
Cite this article
GCRI editorial desk (2026, June 8). Technical trust infrastructure for a more prepared world. The Global Centre for Risk and Innovation. https://therisk.global/the-future-gcri-is-built-for-technical-trust-infrastructure-for-a-more-prepared-world/
@online{gcri2026thefuturegcriisbuilt,
author = {{GCRI editorial desk}},
title = {{Technical trust infrastructure for a more prepared world}},
publisher = {The Global Centre for Risk and Innovation},
year = {2026},
month = {6},
url = {https://therisk.global/the-future-gcri-is-built-for-technical-trust-infrastructure-for-a-more-prepared-world/},
language = {en}
} TY - ELEC AU - GCRI editorial desk TI - Technical trust infrastructure for a more prepared world PB - The Global Centre for Risk and Innovation PY - 2026 DA - 2026/06/08 UR - https://therisk.global/the-future-gcri-is-built-for-technical-trust-infrastructure-for-a-more-prepared-world/ LA - en ER -
Republish
Republish or translate this article free of charge, in print or online, under CC BY 4.0: keep the byline and The Global Centre for Risk and Innovation, link to the original and note any change. Figures, maps and twins made by the institutions carry the same licence; third-party data keeps its own credit line.
Attribution to paste
<p>Technical trust infrastructure for a more prepared world — GCRI editorial desk, The Global Centre for Risk and Innovation. First published 8 June 2026 at <a href="https://therisk.global/the-future-gcri-is-built-for-technical-trust-infrastructure-for-a-more-prepared-world/">https://therisk.global/the-future-gcri-is-built-for-technical-trust-infrastructure-for-a-more-prepared-world/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>