Risk Governance · Technical report

Frontier AI Trends Report

NUWA research note · independent review pending

First published: Publication date not verifiedSource checked: 2026-10-28

Original source ↗Cite this record ↓Autonomy →Deception →Cybersecurity →CBRN →Disinformation →

NUWA Research Note

Research question Design and assumptions Source findings NUWA interpretation Limits and contrary evidence Research and decision relevance Evidence update Institutions and roles Original resources

Model-assisted reading · Selected sections of the original full text or official article; a complete section-by-section audit remains pending.

At a glance

Cross-domain trends require methodology-specific interpretation and negative-evidence limits.

Research question

How should frontier trends be read across distinct evaluation methods?

Design and assumptions

The official report juxtaposes question answering, long tasks, agent evaluation, expert red teaming and human studies, discussing their methodological blind spots. Included studies retain their original experimental identity.

Source findings

The report describes no observable spontaneous sandbagging in more than 2,700 monitored transcripts, while acknowledging black-box limitations. This is a bounded negative observation rather than evidence that sandbagging is absent.

NUWA interpretation

A cross-domain report can organise questions and evidence chains, but methods and detection sensitivity must remain explicit. Non-detection does not establish absence.

Limits and contrary evidence

The report is not an independent replication of its component studies. Black-box and trace-based observations may miss unobservable strategies.

Research and decision relevance

NUWA recommendation: Use its method taxonomy to connect records and identify missing evidence without pooling heterogeneous scores.

Precise source locations

Frontier AI Trends Report · Evaluation methods; sandbagging; scientific capabilities; text lines 00200Original resource captured 2026-10-28T07:24:44.869574+00:00; sha256:b824d2f8d82a6649fb2b1f1eb1420edfecb1e6cc6ebd663210de4d7aab33a012

Source checking, extraction, editorial approval and replication are recorded separately. Methodology →

Evidence update · 2026-10-28

Model-assisted reading · independent evidence and bilingual editorial review pending. Values are source findings or inventory counts, separate from the trend result series.

Download source-bound evidence · JSON ↓

Institutions and roles

Published by UK AI Security Institute

Cite and reuse

Cite the original publication for its findings. Cite this NUWA record separately for its explanation and retained snapshot.

BibTeX ↓RIS ↓Record + source metadata · JSON ↓Reading coverage · JSON Suggest a correction

Original publication · BibTeX

@misc{aisi-trends,
  title = {Frontier AI Trends Report},
  url = {https://www.aisi.gov.uk/frontier-ai-trends-report},
  version = {Original resource captured 2026-10-28T07:24:44.869574+00:00; sha256:b824d2f8d82a6649fb2b1f1eb1420edfecb1e6cc6ebd663210de4d7aab33a012},
  note = {Original resource captured 2026-10-28T07:24:44.869574+00:00; sha256:b824d2f8d82a6649fb2b1f1eb1420edfecb1e6cc6ebd663210de4d7aab33a012}
}

NUWA record · BibTeX

@misc{aisi-trends-nuwa,
  title = {Frontier AI Trends Report — NUWA research record},
  url = {https://whitzard.tech/en/nuwa/whitzard-index/work/aisi-trends/},
  version = {2026-10-29.1},
  organization = {NUWA Frontier AI Safety Lab},
  howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--aisi-trends.json},
  note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--aisi-trends.json.}
}

Evidence snapshot 2026-10-29.1 · Index updates · RSS

References

[1] Frontier AI Trends Report ; Evaluation methods; sandbagging; scientific capabilities; text lines 00200