Public-safe dashboards: showing systemic risk without false authority

A public-safe dashboard is a visual interface that presents risk evidence to people outside the team that produced it, built so that every value carries the information a reader needs to judge it: where it came from, what kind of evidence it is, when it was valid, how uncertain it is, what status it has and how to report an error. In the Nexus Consortium, the public-good programme through which The Global Centre for Risk and Innovation (GCRI), The Global Risks Forum (GRF) and The Global Risks Alliance (GRA) help countries examine interconnected risk with shared methods, the term names a discipline for panels rather than a product ¹.

The discipline is divided among the twenty Nexus systems so that no single tool both draws a panel and decides that it may be shown ². NXS-STU (composition, dashboards and workflows) composes the panel and binds each component to its source references, validating data types, units, permissions and lifecycle compatibility ². NXS-GRX (canonical data, provenance and risk) holds the data products and their lineage ⁹. NXS-NSF (trust, authorization and assurance) issues the public-safe release decision, and NXS-POR (public, client and partner delivery) delivers the panel to its audience with field-level minimisation, embargo and jurisdiction constraints ^10,3.

A public-safe dashboard is therefore not a warning system, a regulatory finding, a rating, an investment signal or a control room. It informs responsible actors; it never acts for them, and it never claims more than the record behind it ^1,8.

A chart compresses a great deal of judgement into a shape. The choice of baseline, the colour scale, the time window and the decision to show or omit a range all shape what a reader concludes, and most readers never see those choices. In systemic risk work the compression is greater still, because a single panel may combine observations, published statistics, model outputs and scenarios from several institutions. A dashboard that looks finished invites trust that the evidence may not yet support.

The failure is rarely dramatic. On 4 October 2020 Public Health England reported that a technical issue in the data load process that transfers positive laboratory results into reporting dashboards had left 15,841 cases recorded between 25 September and 2 October out of the reported daily figures; most of the missing cases were from the most recent days ⁴. The panels looked complete. Nothing on them said that a load had failed.

The lesson the programme draws is that completeness and freshness must be shown, not assumed. A value that stops updating should say so; a missing value should appear as missing, never as zero; and a figure that has been corrected should carry its version and the reason for the change ^1,3.

The first is the source and owner: who released the data and who stands behind the panel. Original source authority is retained, and Nexus estimates stay labelled as estimates; a value drawn from a national statistics office is attributed to it, and the label "published statistics" names a source type, not an endorsement of the finding by that authority ¹.

The second is the status, stated in words. A panel used in an exercise, a scenario or a test says so on its face. The third is the uncertainty: a range or band wherever a value is estimated or modelled, with a note of what the range covers. The fourth is the evidence label, one of nine used across the programme: observed, published statistics, institution-reported, estimated, modelled, forecast, scenario, expert judgement or unknown ¹.

The fifth is time: when the value was valid and when the panel last refreshed. The sixth is the version and correction route: an accepted correction creates a new recorded version, never a silent edit, and anyone may ask for one without membership ¹. NXS-POR keeps the withdrawal, replacement and recipient notification history of each publication, so that a reader who saw an earlier version can learn that it changed ³.

Uncertainty is the element most often left out, because a range makes a panel look less decisive. Operational forecasters have long shown that it can be drawn plainly. The track forecast cone of the US National Hurricane Center is built from circles placed along the forecast track at 12, 24, 36, 48, 60, 72, 96 and 120 hours, each sized so that two-thirds of historical official forecast errors over a five-year sample fall inside it ⁵.

For the 2026 Atlantic season, based on errors from 2021 to 2025, the radius grows from 25 nautical miles at 12 hours to 62 at 48 hours, 95 at 72 hours and 200 at 120 hours; in the Eastern and Central North Pacific it reaches 138 at 120 hours ⁵. The cone describes the probable track of the centre of the storm. Read precisely, it also says that one track error in three falls outside it.

Two rules follow for any risk panel. A range should state its probability, so that a reader knows whether a band covers two-thirds, nine-tenths or all of the plausible outcomes. And a range should state what it is a range of, because a band around a storm track, a flood depth or a loss estimate means something different in each case. The programme’s data products are built to preserve alternative hypotheses and uncertainty, and NXS-GRX forbids flattening observation, inference, prediction and simulation into one unqualified truth value ⁹.

The most serious form of false authority is a panel that looks like an instruction to act. In every jurisdiction, warnings to the public are issued by competent authorities under their own law. The programme does not issue warnings and is not an emergency-dispatch service; public-safety exercises carry clear simulation marking and stop-work controls, and simulation messages are never delivered as official warnings ¹.

The same line runs through the systems. NXS-EWS (multi-hazard monitoring and warning) may prepare a warning candidate in an exercise, but its method boundary forbids labelling a prepared candidate as an official, issued or delivered warning ¹¹. NXS-POR records that a reachable URL proves neither that the intended audience received a message, nor that it understood it, nor that it was officially endorsed ³. A public-safe dashboard may show where a hazard indicator stands; whether to warn is never its decision.

Exercise displays need the same care in the other direction. During the week of controlled exercises in Nexus Universe, nothing simulated is reported as observed impact, and the evidence and outcomes of each mission are published only as agreed for it ^12,13.

A panel that some readers cannot perceive is not public-safe for them. The W3C Web Content Accessibility Guidelines 2.2, a W3C Recommendation since 5 October 2023 (current edition 12 December 2024), require a contrast ratio of at least 4.5:1 for normal text and 3:1 for large text, at least 3:1 for the parts of graphics required to understand the content, and that colour is never the only visual means of conveying information ⁷. About one man in twelve has colour vision deficiency, according to the US National Eye Institute, which is why status in the programme’s design is written in words and never shown by colour alone ¹⁴.

Practice lags the standard. WebAIM’s 2026 analysis of the top million home pages detected WCAG 2 failures on 95.9% of them, up from 94.8% in 2025; because automated tools detect only part of all failures, full conformance was certainly lower than 4.1% ⁶. Low contrast text was the most common failure, on 83.9% of pages, followed by missing alternative text for images at 53.1% and missing form labels at 51% ⁶. Government pages did better than average, with 42.4 detected errors per page against 56.1 overall ⁶.

NXS-STU keeps an accessibility specification for each composition, with keyboard alternatives, focus order and nonvisual outputs, and its method boundary requires that component validation check data and interaction contracts, not only screenshot similarity ². The figures on this page follow the same rule: each has alternative text and a long description, and the data of the charts are tabulated in the text.

Three kinds of evidence appear on this page, and each is read on its own terms. The hurricane cone radii are operational statistics published by the agency that computes them, for one season and two basins ⁵. The accessibility figures are counts from an automated crawl of one million home pages in February 2026, and the issuer itself warns that the absence of detected errors does not show that a page is accessible ⁶. The missing-case episode is a single incident described by the public body responsible, in its own statement ⁴.

None of these is a measurement of the programme. No public-safe dashboard has been published under this design, and no usage or comprehension figures exist for one. The programme is at service level R0, information and enquiries, as the claims and readiness register and programme status record ^8,15.

A panel has a life, and each stage writes a record. It is composed in NXS-STU from admitted components, with its sources bound and its units checked. It is previewed in a profile isolated from live effects and production credentials ². It is labelled. It receives a public-safe release decision from NXS-NSF, the system that holds policy authority; NXS-POR may not create that decision itself ^10,3. It is delivered under a disclosure profile that applies minimisation, embargo and jurisdiction constraints ³.

After release it can be corrected or withdrawn. A correction is a new version with its reason. A delivered panel keeps its withdrawal, replacement and recipient notification history ³. A panel is never simply switched off and forgotten.

The design draws on instruments that already govern public information, and cites them as references, not as claims of conformance. WCAG 2.2 sets the accessibility baseline ⁷. The programme’s own proposed standards add the rest. NS-5, figures, maps and visualization, is the proposed Nexus Standard for figures and maps; NS-1, the UNI and the artefact catalogue, is the proposed Nexus Standard for identifiers ¹⁷.

In the Nexus Risk Standards, NRS-I2 covers risk communication and public-safe dashboards, and NRS-S15 and NRS-S19 are the system profiles of NXS-STU and NXS-POR; all three have status OPEN, which means named and specified but not drafted ^2,3,18. A status such as proposed or open describes a drafting stage only; none records adoption or certification, and each country decides whether to use a standard.

• Generated summaries on panels. Text written by AI models beside a chart can carry more authority than the chart itself; such text needs the same review as the chart it describes.
• Accessibility as a baseline. WCAG 2.2 has been a W3C Recommendation since 5 October 2023 (current edition 12 December 2024), and the share of top home pages with detected failures rose in 2026, from 94.8% to 95.9%, after six years of small improvements, so accessibility review belongs before release, not after ^7,6.
• Probabilistic displays for the public. Operational agencies now publish the statistics behind their uncertainty graphics, as the National Hurricane Center does for its cone each season; risk panels can state their coverage probability in the same way ⁵.
• Freshness as a field. Since a silent load failure can remove thousands of records from a panel, the time of last successful refresh is treated as data in its own right, shown beside the value ^4,9.

GCRI leads scientific and editorial review of public outputs, under its scientific and editorial integrity policy ¹³. GRF stewards Nexus Universe, where exercise displays are used under exercise control, and publishes mission outcomes only as agreed ¹³. GRA, through its Consortium finance interface, reads evidence for the questions competent financial institutions ask; a panel shown in a finance-readiness dialogue informs those questions and carries no promise of funding, underwriting, a bankability rating, capital or a transaction ¹⁹. Each institution sets out its own role on its site: The Global Centre for Risk and Innovation, The Global Risks Forum and The Global Risks Alliance.

For a reader who wants to go further, the companion page on public-safe technical reporting applies the same rules to written findings, and Reports publishes editions only after review.

What this page does not claim: no public-safe dashboard is published under this design, and the programme is at service level R0, information and enquiries. The programme issues no warnings, and a panel is never a warning, a regulatory finding, a rating, an approval or an investment signal. 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, instrument or standard is not a claim of partnership, conformance or endorsement.