GCRI · Reports
Frontier technology risk assessment
Frontier technology reports cover fields that change faster than the publication cycle: artificial intelligence, synthetic biology, quantum computing, advanced materials and space systems. Evidence arrives as preprints, benchmarks, patents and company announcements, and is often superseded within months. This page sets out how fast-moving evidence is assessed, versioned and reviewed, how claims are separated from demonstrated capability, and how information hazards are handled before publication.
Frontier technology risk assessment on the record.
Frontier technology risk assessment: how the risk develops.
Exponentially developing technologies. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Evidence ages quickly
A capability assessment can be outdated by the time it is reviewed. Without dated evidence windows and versioned updates, readers may act on a picture of a technology that no longer holds.
Preprints and announcements
Much frontier evidence is unreviewed: preprints, technical blogs and vendor claims. Repeating such claims without labelling them turns promotion into apparent consensus.
Benchmark fragility
Benchmarks saturate, leak into training data or measure narrow skills. Reports that equate benchmark scores with real-world capability overstate readiness and understate failure modes.
Information hazards
Some technical detail could enable harm if published. Pre-publication dual-use review weighs the value of disclosure against misuse, and records what was withheld and why without exposing it.
Series production desk.
The screen a team would use to follow this subject: its measures, states and decisions. Names are examples and values are illustrative.
Series production desk
Frontier technology risk assessment · Annual synthesis · Edition 3
Editions in production
| Annual synthesis | Global | Data check | 83 | W18 |
| Country profile 7 | Country | Peer review | 98 | W38 |
| Water outlook | Sectoral | Drafting | 46 | W13 |
| Methods note 4 | Methods | Released | 100 | done |
Data check, annual synthesis
Edition pipeline
- Outlineboard
- Draftauthors
- Data checkevery value
- Peer reviewtwo rounds
- Editorialdecision
- ReleaseDOI
- Correctionsversioned
Technology editions trace effects into each domain.
Editions separate demonstrated results from projections. Models discussed carry NS-12 cards under the proposed Nexus Standards.
- Status updates with sources.
- Corrections to claims.
How frontier technology risk assessment is modelled, measured and tested.
The model, its assumptions, its data and its uncertainty, in the form a scientist or engineer would check. Structures follow published methods. The values drawn are illustrative, not findings.
Model, equations and assumptions
The model structure used for frontier technology risk assessment, the assumptions it makes and how each is tested.
Ri = f(Hi, Ei, Vi, Ci)hazard, exposure, vulnerability and coping capacityEAL = ∑k pk · Lkover events k with annual probability pkxt+1 = A xt + B ut + εtA carries the couplings, u the shock, ε the error| Assumption | Why it is made | How it is tested |
|---|---|---|
| Hazard and exposure are independent within an event | Keeps the loss integral tractable | Compare with a copula fit on joint records |
| Couplings are linear near the operating point | Lets shocks propagate through A | Perturb beyond the linear range and compare |
| Vulnerability curves transfer between places | Local curves are rarely available | Validate against observed losses where they exist |
- Inputs
- versioned datasets with provenance
- Outputs
- distributions, never single numbers
- Code
- released with the model card and tests
- Reproduction
- same data, code, environment and seed
What is measured, and how well.
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Preprint share of cited evidence | percent of references | per version | Reference metadata | OBSERVED |
| Median age of cited evidence at publication | months | per version | Reference metadata | OBSERVED |
| Interval between report versions | months | per version | Version history | Published statistics |
| Claims labelled by demonstration status | percent of capability statements | per version | Editorial audit | REPORTED |
| Cited preprints later published after review | percent of cited preprints | quarterly | Preprint to publication links | OBSERVED |
| Dual-use review completed | yes or no, with reviewer role | per version | Review record | DECLARED |
Where the evidence comes from.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| arXiv | Preprints in computer science, physics, quantitative biology and related fields | Metadata CC0, licence per paper |
| bioRxiv and medRxiv | Preprints in the life and health sciences with links to later publications | Licence per preprint |
| OpenAlex | Publication, citation and preprint-to-article links | CC0 |
| WIPO PATENTSCOPE | International and national patent applications for technology tracking | Free to search, terms per WIPO |
| The Lens | Linked scholarly and patent records | Free to use, terms per Lens |
| OECD.AI Policy Observatory | National AI policies, initiatives and indicators | Free to use, terms per OECD |
Who decides, and what the evidence serves.
Reports prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Science, technology and innovation ministries | Which technologies need governance attention, investment or monitoring now | Claim registers by demonstration status, with dated evidence windows |
| Standards bodies and technical committees | Where standards for testing, safety or interoperability are needed | Documented gaps between claimed and demonstrated capability |
| Research funders and security review panels | Whether research areas carry dual-use concerns that need oversight | Dual-use review records and the categories of hazard identified |
| Journal and series editors | When a report must be updated or marked as superseded | Evidence age, preprint status changes and version intervals |
Outputs this topic produces
- Versioned technology reports with dated evidence windows
- Claim registers separating demonstrated, reported and projected capability
- Living updates that show what changed between versions
- Dual-use review records stating the basis for any withheld detail
What the series asks.
Fast-moving technologies outpace the evidence and institutions that govern them. Claims about their capabilities and risks range from careful to wildly speculative. Readers need analysis that separates demonstrated capability from projection, explains the mechanisms by which a technology could create or reduce risk, and states how quickly conclusions may age.
- What can the technology demonstrably do today, under what conditions, and what is still projection?
- Through which mechanisms could it create new risks, such as misuse, failure, concentration or loss of control?
- How could it reduce existing risks, and what would need to be true for that benefit to materialise?
- Which governance, safety and security measures are being developed, and what evidence exists on their effect?
What the series covers.
The series reads technology through capability, risk and governance.
Demonstrated versus claimed
Evidence of performance is labelled by source; vendor claims are labelled Institution-reported.
Mechanisms of harm
Accidents, misuse, systemic dependence and security effects, analysed with the relevant specialist platforms.
Responses
Standards, safety practice and policy options, described without endorsing any product.
Technology analysis is most useful when it follows mechanisms rather than headlines. A general-purpose technology changes risk through the systems that adopt it: an AI model embedded in grid operations, a quantum computer able to break widely used public-key cryptography, a biotechnology tool that lowers the barrier to working with pathogens. Each edition therefore pairs the technology with the specialist platforms that study the affected systems, and names the conditions under which its conclusions would change.
How editions are built, and their limits.
- 01
Technical literature
Peer-reviewed and preprint research, with preprints labelled as such.
- 02
Benchmarks and tests
Published benchmarks and tests, with their limits; Labs results appear only from completed protocols.
- 03
Expert elicitation
Structured judgement on timelines and impacts, labelled Expert judgement with its panel described.
- 04
Shelf life
Each edition states when its findings should be reviewed, because technical facts change quickly.
How it is studied, and what the methods cannot say.
- 01Uncertainty quantification and model validationCalibrated language in assessments · Certainty of evidence ratingsApplied here
- 02Data engineering, interoperability and stewardshipStructured publishing in JATS XML · Persistent identifiers across the recordApplied here
- 03Probability and statisticsMeta-analysis with heterogeneity stated · Small-study and publication biasApplied here
Where this page’s references lead.
The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.
Questions and answers.
Why are preprints cited if they are not reviewed?
What is a living report?
What does dual-use review decide?
Does the series endorse technologies or vendors?
How does the series handle hype?
Are dual-use risks discussed openly?
How quickly are editions updated?
Contribute evidence from this page.
Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
- EmailGateway GW-GCRISend it from this pageEvidence, a correction, a case, a method, a review, a translation or profile input. It arrives tagged with this page, so the editor who owns it receives it.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsGCRISee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Never send personal data, sensitive infrastructure detail or anything you have no right to share. A contribution is reviewed before use; sending it confers no authorship or editorial rights, and sponsorship never buys a finding.
Bring a method, a dataset or a question.
Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Legal documents.
Maps, data and figures on this site are shown for context and general information. They are not advice, forecasts or warnings; the limits of the information are set out in the Disclaimers and Data Notices.