GCRI · Foundations
Claims discipline: how GCRI keeps technical claims within the evidence
Systemic-risk evidence is used by people who cannot check it themselves, so an overstated claim becomes a risk of its own. This page explains the claims discipline of The Global Centre for Risk and Innovation (GCRI): states from a record, eight separations, nine evidence labels and an open correction route.
- As of
- 21 October 2026
- 読む
- 11 minutes
- エビデンス
- 14 sources, cited in order
手法とエビデンス
- エビデンス
- 14 sources, cited in order and listed under References
- 数字
- 4 figures
- Data availability
- Every chart's data is published as a table on this page.
01What claims discipline is
What claims discipline is
Claims discipline is the practice of matching every public statement to the record behind it: a programme's state, a system's maturity, a finding's evidence and an institution's role are described as recorded, with their limits, and changed only when the record changes. In the Nexus Consortium, the public-good programme of The Global Centre for Risk and Innovation (GCRI), The Global Risks Forum (GRF) and The Global Risks Alliance (GRA), the rule is stated in one line: "every programme carries one explicit state, taken from a record, never from ambition" 1.
Governance gives the rule its teeth. It decides who may change a state, what evidence the change needs, who may challenge it and how a wrong statement is corrected. This page explains that machinery as GCRI applies it, from the claims and readiness register to the correction route, and why technical institutions that work on systemic risk need it more than most.
02Why overclaim is a systemic risk of its own
Why overclaim is a systemic risk of its own
Systemic-risk work is used by people who cannot check it themselves: a finance ministry reading a resilience estimate, a utility reading a model of cascading failure, an insurer reading an exposure map. An overstated claim travels through those readers into decisions, budgets and contracts, and when it fails it takes the credibility of the honest work with it. That is a cascade in its own right.
Regulators have begun to price it. On 18 March 2024 the US Securities and Exchange Commission charged two investment advisers with false and misleading statements about their use of artificial intelligence and imposed civil penalties of US$225,000 and US$175,000, US$400,000 in total 3. The EU directive "on empowering consumers for the green transition", adopted on 28 February 2024, targets generic environmental claims made without substantiation and claims that rely on offsetting; member states had to transpose it by 27 March 2026, and its rules apply from 27 September 2026 4. The UK Financial Conduct Authority's anti-greenwashing rule, in force since 31 May 2024, requires any reference to a product's sustainability characteristics to be consistent with them and "fair, clear and not misleading" 7,8.
Science is under the same pressure. Nature reported that more than 10,000 research papers were retracted in 2023, a record 9. Public trust holds, but on specific grounds: a 68-country survey of 71,922 people found trust in scientists moderately high, with a mean of 3.62 on a five-point scale, highest for competence (4.02) and lowest for openness to feedback (3.33) 5. Institutions keep trust by being exact about what they know and open about what they do not.
Fig. 01Regulators and publishers are pricing overclaim
Figure 1 in words
A horizontal timeline from late 2023 to 2027 with six events: December 2023, Nature reports more than 10,000 research papers retracted in 2023, a record; 28 February 2024, the EU adopts Directive 2024/825, "on empowering consumers for the green transition"; 18 March 2024, the US SEC charges two investment advisers over false AI claims, with 400,000 US dollars in penalties; 31 May 2024, the UK FCA anti-greenwashing rule ESG 4.3.1R and guidance FG24/3 come into force; 27 March 2026, deadline for EU member states to transpose the directive; 27 September 2026, the directive's rules on generic environmental claims apply.
03Evidence is not authority
Evidence is not authority
The first discipline is to keep evidence and authority apart. The programme's decision rights say so directly: scientific and editorial acceptance belongs to qualified, conflict-managed reviewers and the publisher, and "institutional influence cannot decide a scientific conclusion"; decisions and their execution belong to the relevant government, operator or licensed institution 10. The Decide step of the Nexus cycle sits outside the programme for that reason.
It follows that no output of the programme speaks for an authority. A country profile never implies a government commission where none exists; a city relationship never becomes national representation; regulator participation confers no regulatory approval or privileged access; and agencies use their own authorised instruments, so participation creates no endorsement by the UN or any agency 2. GCRI does not regulate or certify, GRF does not regulate, finance or procure, and GRA does not lend, underwrite, rate or invest.
04States from a record, never from ambition
States from a record, never from ambition
Every programme has one of seven states: exploration, in formation, constituted, active programme, paused, closed or historical record 2. A state moves only when the record behind it changes; the accountable structure records the change with its evidence, and the Secretariat publishes it 2. Programmes can move backwards, and a closed programme stays visible as a historical record. On 21 October 2026 all six programme regions are in formation, three named consortiums are in formation, the fifteen planned hubs are coordination roles with none operating, and no national programme is recorded as constituted or active 2.
Service is graded the same way, from R0 information to R5 public release, and each level opens only when its service team, terms and review capacity are in place 2. The current level is R0: enquiries go to regional and country desks and receive a reference, and applications, payments, computing allocations and registrations are not yet open 6,2.
Fig. 02Six levels of service, one open today
Figure 2 in words
Six rising steps. R0 Information: visitors can understand the architecture and current programme state; highlighted as the current level on 21 October 2026. R1 Enquiry: submit a question and receive a durable receipt from a staffed queue. R2 Application: apply for a defined pathway with current terms and an appeal route. R3 Commitment: an accepted participant enters an agreement with the correct parties and scope. R4 Operation: authorised users undertake the scoped service, workload or exercise. R5 Public release: evidence or public status is released and maintained with independent release authority.
05Eight separations that stop a step passing for the next
Eight separations that stop a step passing for the next
Most overclaim is not invention but promotion: a real step described as the one after it. The model lists the promotions it forbids 2. An application is not membership; a claimed identity is not representative authority; a payment is not an appointment or access to restricted data; an infrastructure offer is not an allocation; a report candidate is not a released publication; a research coverage profile is not a constituted consortium or local office; a planned hub is not an operational site; and readiness evidence is not a financing or underwriting decision.
Each separation guards a real temptation. A sponsor would like a contribution to read as a seat; a provider would like an offer to read as a deployment; a city would like a profile to read as a partnership. The separations make the honest description the only permitted one, and they apply to the programme's own institutions as strictly as to anyone else.
Fig. 03Eight things that are never the next state
Figure 3 in words
Eight rows, each pairing a step with what it is never treated as: a submitted application, not membership; a claimed identity, not representative authority; a payment, not an appointment or access to restricted data; an infrastructure offer, not an allocation; a report candidate, not a released publication; a research coverage profile, not a constituted consortium or local office; a planned hub, not an operational site; readiness evidence, not a financing or underwriting decision. A note states that a state changes only when the record behind it changes, recorded by the accountable structure and published by the Secretariat.
06Claims by kind: authorities, sponsors, providers, finance and AI
Claims by kind: authorities, sponsors, providers, finance and AI
Public authorities. Naming an agency, ministry or regulator is not a claim of partnership. A relationship is described only as recorded, and the authority's own decision is never implied by programme work 2,10.
Sponsors and providers. No support buys appointments, findings, consent or procurement advantage; sponsorship never influences a testbed result; and testbed participation is not a regulatory sandbox, a certification or a procurement status 2. Offered capacity is never counted as reserved, consumed or funded 2.
Finance and insurance. GRA's readiness work produces diligence questions, not decisions: there is no promise of funding, underwriting, a bankability rating, capital or a transaction, and a readiness record is not loan approval, a grant award or procurement qualification 2. The boundary is set out in lawful handoffs and GRA's non-execution boundary.
Models, simulations and AI. The model pages carry four kinds of number, teaching values, illustrative applications, published benchmarks and specifications, and each is labelled so that none is read as a measurement or a forecast 11. Simulated results are never reported as observed impact 2. Model and data cards follow the proposed standard NS-12 for software, models and AI artefacts, and the use of AI in GCRI's own publishing is disclosed under its AI-use statement.
Communities. Protected knowledge never enters public examples or model training, and no membership or account is needed to raise a harm or correct public information 2.
07Nine labels for every finding
Nine labels for every finding
Every finding the programme publishes carries one of nine evidence labels: observed, published statistics, institution-reported, estimated, modelled, forecast, scenario, expert judgement or unknown 1,2. The label "published statistics" names a source type, data reported by an authority; it is not an endorsement of a finding by that authority. Missing data is shown as missing, never as zero 2. The labels are what let a reader see at a glance whether a number was measured, reported, modelled or guessed, and they travel with the number into every edition of Reports.
Participation states are kept in the same spirit, in the programme's participation records, with evidence classes documented for each kind of record. Recognition follows the record, never the claim.
Fig. 04Trust in scientists is moderately high, and lowest on openness
Figure 4 in words, and its data
A dot chart with a horizontal axis from 1, very low, to 5, very high, with 3 marked neither high nor low. Overall trust in scientists 3.62; competence 4.02; integrity 3.58; benevolence 3.55; openness to feedback 3.33. Data from Cologna et al., Nature Human Behaviour, 2025, a preregistered survey of 71,922 respondents in 68 countries.
| Dimension | Mean rating | 出典 |
|---|---|---|
| Overall trust in scientists | 3.62 5 | Cologna et al. 2025 |
| コンピテンス | 4.02 5 | Cologna et al. 2025 |
| 完全性 | 3.58 5 | Cologna et al. 2025 |
| Benevolence | 3.55 5 | Cologna et al. 2025 |
| Openness to feedback | 3.33 5 | Cologna et al. 2025 |
08Claims review and the correction of overclaim
Claims review and the correction of overclaim
Claims are reviewed where they are made. Before an edition is published, series editors and independent reviewers check that each finding carries its label, method and limits; no funder, sponsor or member can buy or block a finding, as the editorial governance page sets out. Programme states are reviewed against the record before the Secretariat publishes them, and GCRI's claims and readiness register, GRF's claims and programme state and GRA's claims and programme status publish the same states on each site 6,12,13.
When a claim is wrong, anyone may ask for a correction, without membership 2. The corrections route runs from report to acknowledgement, review, decision and record, and an accepted correction creates a new recorded version; histories stay visible, and edits are never silent 14. A programme or decision complaint is reviewed by someone not responsible for the contested decision 2.
09Standards and instruments for claims
Standards and instruments for claims
The programme's own rules sit alongside the external instruments that now govern technical and environmental claims in several jurisdictions.
| インストゥルメント | What it requires | ステータス |
|---|---|---|
| NRS-I2 | リスクコミュニケーションと一般公開向けの安全なダッシュボード | Proposed Nexus Risk Standard |
| NRS-I3 | リスクのためのモデルカードとデータカード | Proposed Nexus Risk Standard |
| NS-10 | 方針と条項 | Proposed foundation standard |
| Directive (EU) 2024/825 | No generic environmental claims without substantiation; offsetting-based claims addressed | Applies from 27 September 2026 4 |
| FCA ESG 4.3.1R | Sustainability references must be fair, clear and not misleading | In force since 31 May 2024 7,8 |
| SEC enforcement on AI claims | Statements about AI use must be true and not misleading | Settled actions, 18 March 2024 3 |
10What is changing
What is changing
- AI claims are an enforcement priority. US actions over false statements about AI use reached settled penalties in 2024 3.
- Environmental claims need substantiation by default. EU rules on generic environmental claims now apply across the single market 4.
- Retraction is rising. A record of more than 10,000 retracted papers in 2023 shows that correction systems are being tested at scale 9.
- Openness is the weak point of trust. Scientists score lowest on openness to feedback, which is what correction routes and visible histories address 5.
11One discipline across three institutions
One discipline across three institutions
GCRI applies claims discipline to evidence, methods and reports; GRF to councils, convening and Nexus Universe, where what membership means is set out and recognition records explain why trust must be built by evidence; GRA to finance readiness, where membership is not authority. The companion GCRI essay on claims discipline in shared resilience infrastructure treats the same rules from the side of systems and data, and the public-safe language guide applies them to campaigns. Each institution's mandate is on its home page: GCRI, GRF and GRA.
What this page does not claim: no Nexus programme is recorded as constituted or active, every region is in formation and the service level is R0. GCRI does not regulate or certify, GRF does not regulate, finance or procure, and GRA does not lend, underwrite, rate or invest. Citing a regulator's rule or action describes that rule or action; it is not a claim that any programme activity falls under it or has been reviewed by that regulator. Naming an agency or organisation is not a claim of partnership or endorsement.
質問
質問と回答
Who decides what state a programme is in?
The accountable structure records a change of state with its evidence, and the Consortium Secretariat publishes it. A state is never upgraded for publicity, and it moves only when the record behind it changes.
Can a sponsor or member say it is a partner of GCRI?
Only where a recorded agreement says so, and only in the terms it records. No support buys an appointment, a finding, consent or procurement advantage, and a payment is never an appointment or access to restricted data.
Does a Reports edition carry the authority of a government?
No. An edition is published after review and is never a government risk register or a sovereign rating. The label published statistics names a source type, not an endorsement of a finding by the authority that reported the data.
How do I correct a claim I think is wrong?
Use the corrections and concerns route on any of the three sites. It is open to anyone without membership, and an accepted correction creates a new recorded version with the change and its reason visible.
Why does the programme publish that it is only at service level R0?
Because a register that only ever reports progress is not a register. Stating the current level, and what is not yet open, is the claim the record supports today.
利用条件
用語集
- Claims discipline
- Matching every public statement to the record behind it, with its limits, and changing it only when the record changes.
- プログラムの状態
- One of seven recorded states, from exploration to historical record, that every programme carries.
- Separation
- A rule that one step, such as an application or an offer, is never treated as the next state.
- エビデンス・ラベル
- One of nine labels on every finding, from observed to unknown.
- サービスレベル
- One of six levels, R0 to R5, describing what participants can do; R0 today.
- Overclaim
- A statement that promotes a step, a result or a role beyond what the record supports.
Reuse and cite
Text CC BY 4.0. Cite the page with its date; data tables are linked under each figure.
出典
参考文献
14 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.
- Nexus documentation (docs.therisk.global) (2026-10-21). Status and service levels. docs.therisk.global. Accessed 2026-10-21.
- US Securities and Exchange Commission (2024-03-18). SEC charges two investment advisers with making false and misleading statements about their use of artificial intelligence (press release 2024-36). www.sec.gov. Accessed 2026-10-21.
- European Commission (2024-02-28). Directive (EU) 2024/825, "on empowering consumers for the green transition". transition-pathways.europa.eu. Accessed 2026-10-21.
- Cologna et al., Nature Human Behaviour 9, 713 to 730 (2025-01-20). Trust in scientists and their role in society across 68 countries. www.nature.com. Accessed 2026-10-21.
- The Global Centre for Risk and Innovation (2026-10). Claims and readiness register. therisk.global. Accessed 2026-10-21.
- UK Financial Conduct Authority, FCA Handbook (2024-05-31). ESG 4.3 Anti-greenwashing rule (ESG 4.3.1R). www.handbook.fca.org.uk. Accessed 2026-10-21.
- UK Financial Conduct Authority (2024-04). FG24/3: Finalised non-handbook guidance on the Anti-Greenwashing Rule. www.fca.org.uk. Accessed 2026-10-21.
- Van Noorden, Nature 624 (2023-12-12). More than 10,000 research papers were retracted in 2023, a new record. www.nature.com. Accessed 2026-10-21.
- Nexus documentation (docs.therisk.global) (2026-10-21). Decision rights. docs.therisk.global. Accessed 2026-10-21.
- Nexus documentation (docs.therisk.global) (2026-10-21). Method and limits. docs.therisk.global. Accessed 2026-10-21.
- The Global Risks Forum (2026-10). Claims and programme state. globalriskforum.com. Accessed 2026-10-21.
- The Global Risks Alliance (2026-10). Claims and programme status. globalriskalliance.com. Accessed 2026-10-21.
- The Global Centre for Risk and Innovation (2026-10). Corrections and concerns. therisk.global. Accessed 2026-10-21.
記録
Editorial record
Facts and programme states are given as of 21 October 2026. This page replaces an earlier version on governance and claims discipline; it keeps that version's argument and replaces its descriptions of recognition and archive governance with the states, separations, labels and correction routes now on record. It is reviewed at least every 180 days and whenever a programme state, a cited rule or a cited figure changes.
Published by the GCRI editorial desk under the scientific and editorial integrity policy and the independence and conflicts policy. To report an error or request a correction, use corrections and concerns; corrections are dated and listed with the page.
Cite this article
GCRI editorial desk (2026, June 8). Claims discipline: keeping technical claims within evidence. The Global Centre for Risk and Innovation. https://therisk.global/governance-and-claims-discipline-how-gcri-protects-technical-trust-from-overclaim/
@online{gcri2026governanceandclaimsdi,
author = {{GCRI editorial desk}},
title = {{Claims discipline: keeping technical claims within evidence}},
publisher = {The Global Centre for Risk and Innovation},
year = {2026},
month = {6},
url = {https://therisk.global/governance-and-claims-discipline-how-gcri-protects-technical-trust-from-overclaim/},
language = {en}
} TY - ELEC AU - GCRI editorial desk TI - Claims discipline: keeping technical claims within evidence PB - The Global Centre for Risk and Innovation PY - 2026 DA - 2026/06/08 UR - https://therisk.global/governance-and-claims-discipline-how-gcri-protects-technical-trust-from-overclaim/ 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>Claims discipline: how GCRI keeps technical claims within the evidence — GCRI editorial desk, The Global Centre for Risk and Innovation. First published 8 June 2026 at <a href="https://therisk.global/governance-and-claims-discipline-how-gcri-protects-technical-trust-from-overclaim/">https://therisk.global/governance-and-claims-discipline-how-gcri-protects-technical-trust-from-overclaim/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>