The Global Centre for Risk and Innovation (GCRI)
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 Evidence arrives as preprints, benchmarks, patents and company announcements, and is often superseded within months.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 reports cover fields that change faster than the publication cycle, such as artificial intelligence, synthetic biology, quantum computing, advanced materials and space systems. Evidence arrives as preprints, benchmarks, patents and announcements and ages within months, so claims carry their demonstration status and reports are versioned.
The working screen behind this subject: its structure, terms and states. Names are placeholders and values illustrative.
Frontier technology series · Annual synthesis · Edition 3
| 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 |
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.
A capability assessment can be outdated by the time it is reviewed.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.
Much frontier evidence is unreviewed: preprints, technical blogs and vendor claims. Repeating such claims without labelling them turns promotion into apparent consensus.
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.
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.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Preprint | A version of a paper shared publicly before peer review. |
| Demonstration status | Whether a claimed capability has been shown in a lab, in a pilot, in use, or only announced. |
| Information hazard | A risk that arises from disseminating true information that could enable harm. |
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 | OFFICIAL |
| 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 |
The topic as a computational graph: named sources feed the methods, an evidence review gates the outputs, and the decisions stay with the authorities that hold them.
Fig. 02Nodes carry their type, two metric rows and a state; connectors show what feeds what; dashed frames name the owners; the timeline is the cadence. Design of the pipeline, not a record of a run.
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 |
Nexus 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 |
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.
The series reads technology through capability, risk and governance.
Evidence of performance is labelled by source; vendor claims are labelled Institution-reported.
Accidents, misuse, systemic dependence and security effects, analysed with the relevant specialist platforms.
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.
Peer-reviewed and preprint research, with preprints labelled as such.
Published benchmarks and tests, with their limits; Labs results appear only from completed protocols.
Structured judgement on timelines and impacts, labelled Expert judgement with its panel described.
Each edition states when its findings should be reviewed, because technical facts change quickly.
The current service level is R0: information and enquiries. 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.