RiskNet: A large-scale dataset of AI risk incidents from news with alignment and multi-dimensional annotations

Haochuan Liu; Jinliang Chen; Leihan Zhang; Qiang Yan; Qianyu Zhang; Wecheng Ye; Xianlong Ma; Yang Li

arxiv: 2606.08376 · v1 · pith:MOLIP27Enew · submitted 2026-06-07 · 💻 cs.LG · cs.AI

RiskNet: A large-scale dataset of AI risk incidents from news with alignment and multi-dimensional annotations

Leihan Zhang , Wecheng Ye , Xianlong Ma , Haochuan Liu , Yang Li , Qianyu Zhang , Jinliang Chen , Qiang Yan This is my paper

Pith reviewed 2026-06-27 18:35 UTC · model grok-4.3

classification 💻 cs.LG cs.AI

keywords AI risk incidentsnews datasetincident alignmentrisk labelingmultilingual sourcesbenchmark datasetsAI safetygovernance

0 comments

The pith

RiskNet turns large-scale multilingual news into aligned clusters of AI risk incidents with multi-dimensional labels.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents RiskNet as a dataset built from hundreds of millions of news records to organize scattered reports of AI harms into incident-centered records. It does this through a pipeline that identifies relevant news, screens reports at the event level, aligns related reports into incidents, and applies multi-dimensional risk classifications. The result includes benchmark subsets for tasks like event classification and incident alignment. A sympathetic reader would care because the resource aims to support ongoing, data-driven tracking of real-world AI failures where high-level governance principles currently lack detailed empirical backing.

Core claim

The central claim is that a structured pipeline for AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification applied to large-scale multilingual news sources yields a resource that organizes dispersed reports into incident-centered records, including aligned incident clusters and annotated benchmark subsets, thereby enabling downstream research on AI safety, governance, risk analysis, and longitudinal analyses of harms.

What carries the argument

The structured pipeline for AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification that converts raw news into incident-centered records and benchmark datasets.

Load-bearing premise

The structured pipeline for AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification produces accurate, representative, and unbiased incident records from news sources.

What would settle it

An independent audit that finds a substantial fraction of the aligned incident clusters merge unrelated events or systematically omit documented AI risk cases from the source news.

read the original abstract

As artificial intelligence (AI) systems are increasingly deployed across socially consequential domains, reports of AI-related harms and failures have grown in frequency and diversity. Although existing governance frameworks articulate high-level principles for responsible AI, large-scale empirical resources for tracking and analyzing real-world AI risk incidents remain limited. Existing incident collections are often manually curated, relatively small in scale, and insufficient for continuous, data-driven monitoring and downstream computational analysis. To address this need, we present RiskNet, a large-scale dataset of AI risk incidents constructed from large-scale multilingual news sources. RiskNet applies a structured pipeline for AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification. The resulting resource organizes dispersed news reports into incident-centered records and provides benchmark datasets for event classification, incident alignment, and incident-level risk labeling. In its current release, RiskNet covers hundreds of millions of source records and yields a large-scale collection of AI risk-related reports, including aligned incident clusters and annotated benchmark subsets. The dataset is also accessible through an online platform for browsing and exploration. We describe the data sources, processing workflow, taxonomy design, and technical validation of the resource. RiskNet is intended to support downstream research on AI safety, governance, risk analysis, and benchmarking, as well as longitudinal and cross-source analyses of AI-related harms. By providing a structured and reusable empirical resource, RiskNet helps bridge the gap between high-level governance principles and the documented realities of AI risk incidents.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

RiskNet scales up AI incident collection from news but the abstract gives no evidence the pipeline actually works.

read the letter

The core fact is that this paper releases a dataset built by scraping multilingual news for AI risk reports, clustering them into incidents, and adding labels across several dimensions. That is the main offering.

What they actually did is run a pipeline that first filters news for relevance, screens individual reports, aligns reports that describe the same incident, and then annotates the resulting incidents on risk type and other axes. They also release benchmark subsets for three tasks and put up a browsing platform. The scale claim—hundreds of millions of source records processed down to a large collection of incidents—sounds bigger than the usual hand-curated lists that top out at a few hundred entries.

The obvious gap is that none of the validation numbers appear in the abstract. There are no accuracy figures, no inter-annotator agreement, no error analysis on the alignment step, and no check on whether the news sources over- or under-represent certain kinds of incidents. The paper says technical validation is described, but without those results the central promise (accurate, representative incident records) stays untested. That is the load-bearing assumption.

This is the kind of resource that AI governance and safety groups might actually use for longitudinal tracking or as training data for event classifiers. A reader who needs raw counts of reported incidents or wants to run their own analyses on the released benchmarks could get something out of it, provided the quality checks hold up in the full text.

I would send it to review. The dataset direction is worth referee time even if the current version needs a clear validation section added.

Referee Report

2 major / 1 minor

Summary. The manuscript presents RiskNet, a large-scale dataset of AI risk incidents constructed from multilingual news sources using a structured pipeline for AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification. The resource organizes dispersed news reports into incident-centered records and provides benchmark datasets for event classification, incident alignment, and incident-level risk labeling, along with an online platform. The paper describes the data sources, processing workflow, taxonomy design, and technical validation of the resource to support AI safety, governance, and risk analysis research.

Significance. If the pipeline is validated to produce accurate and representative records, RiskNet would be a significant contribution by supplying a large-scale empirical resource (from hundreds of millions of source records) that addresses the limitations of small, manually curated incident collections. This could enable new data-driven, longitudinal, and cross-source analyses in AI risk monitoring and benchmarking.

major comments (2)

[Technical validation section] Technical validation section (as referenced in the abstract and processing workflow description): The manuscript states that it describes the technical validation but supplies no quantitative metrics such as precision/recall for news identification, inter-annotator agreement for annotations or alignment, or error analysis for any pipeline stage. This is load-bearing for the central claim that the dataset yields accurate incident records.
[Processing workflow] Processing workflow and pipeline description: The claim that the structured pipeline produces representative and unbiased incident records from news sources lacks supporting evidence such as bias mitigation steps, comparison to existing curated datasets, or validation against ground-truth incidents; without this, the utility for downstream tasks cannot be assessed.

minor comments (1)

[Abstract] Abstract: Mention of 'technical validation' would be strengthened by previewing at least one key quantitative result to allow immediate assessment of the pipeline's performance.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for highlighting the need for stronger empirical support in the technical validation and processing workflow sections. We address each major comment below and commit to revisions that strengthen the manuscript without overstating current content.

read point-by-point responses

Referee: [Technical validation section] The manuscript states that it describes the technical validation but supplies no quantitative metrics such as precision/recall for news identification, inter-annotator agreement for annotations or alignment, or error analysis for any pipeline stage. This is load-bearing for the central claim that the dataset yields accurate incident records.

Authors: We agree that the current manuscript provides only a qualitative description of the validation process without numerical metrics. In the revised version we will add precision/recall estimates for the news identification stage, inter-annotator agreement scores for annotation and alignment tasks, and an error analysis for each pipeline component to directly support the accuracy claims. revision: yes
Referee: [Processing workflow] The claim that the structured pipeline produces representative and unbiased incident records from news sources lacks supporting evidence such as bias mitigation steps, comparison to existing curated datasets, or validation against ground-truth incidents; without this, the utility for downstream tasks cannot be assessed.

Authors: We acknowledge the absence of explicit bias mitigation details and external comparisons in the current text. The revised manuscript will expand the workflow section to document source diversity measures, filtering criteria, and a comparison against existing curated AI incident collections. Direct validation against comprehensive ground-truth incident lists remains constrained by the lack of such external references, which we will note as a limitation while emphasizing the utility of the released benchmark subsets. revision: partial

standing simulated objections not resolved

Direct validation against comprehensive ground-truth incident records, as no complete external ground-truth collection of all AI risk incidents exists for comparison.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes construction of a news-derived dataset via an identification, screening, alignment, and classification pipeline. No equations, parameter fitting, predictions, or first-principles derivations are claimed. The contribution is empirical resource creation with technical validation described; no load-bearing step reduces to self-definition, fitted inputs renamed as predictions, or self-citation chains. This matches the default expectation for non-derivational dataset papers.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The dataset rests on the assumption that news sources reliably surface AI risk incidents and that the automated pipeline can accurately identify, align, and label them; these are domain assumptions without independent verification in the abstract.

axioms (1)

domain assumption Multilingual news sources contain sufficient and representative coverage of AI risk incidents
The entire construction begins from news records and treats them as the primary signal for incident detection.

pith-pipeline@v0.9.1-grok · 5822 in / 1213 out tokens · 47570 ms · 2026-06-27T18:35:27.717785+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

33 extracted references · 13 canonical work pages

[1]

Technical report, Depart- ment for Science, Innovation and Technology (UK) (2026)

Bengio, Y., et al.: International ai safety report 2026. Technical report, Depart- ment for Science, Innovation and Technology (UK) (2026). A global synthesis of scientific evidence on GPAI risks. https://arxiv.org/abs/2602.21012

arXiv 2026
[2]

Managing extreme AI risks amid rapid progress

Bengio, Y., Hinton, G., Yao, A., Song, D., Abbeel, P., Darrell, T., Harari, Y.N., Zhang, Y.-Q., Xue, L., Shalev-Shwartz, S., Hadfield, G., Clune, J., Maharaj, T., Hutter, F., Baydin, A.G., McIlraith, S., Gao, Q., Acharya, A., Krueger, D., Dragan, A., Torr, P., Russell, S., Kahneman, D., Brauner, J., Mindermann, S.: Managing extreme AI risks amid rapid pro...

work page doi:10.1126/science.adn0117 2024
[3]

Winfield, A.F.T., Jirotka, M.: Ethical governance is essential to building trust in robotics and artificial intelligence systems. Philosophical Trans- actions of the Royal Society A: Mathematical, Physical and Engineering Sciences376(2133), 20180085 (2018) https://doi.org/10.1098/rsta.2018.0085 https://royalsocietypublishing.org/doi/pdf/10.1098/rsta.2018.0085

work page doi:10.1098/rsta.2018.0085 2018
[4]

Big Data & Society , author =

Mittelstadt, B.D., Allo, P., Taddeo, M., Wachter, S., Floridi, L.: The ethics of algorithms: Mapping the debate. Big Data & Society3(2), 2053951716679679 (2016) https://doi.org/10.1177/2053951716679679 https://doi.org/10.1177/2053951716679679

work page doi:10.1177/2053951716679679 2016
[5]

TIME Magazine (2025)

Atherton, K.D.: Reports of ai-related incidents rose 50% year-over-year: The new kinds of ai harm emerging. TIME Magazine (2025)

2025
[6]

Staufer, L., Casper, S., et al.: The 2025 AI Agent Index: Documenting Technical and Safety Features of Deployed Agentic AI Systems (2026)

2025
[7]

Technical Report NIST AI 100-1, National Institute of Standards and Technol- ogy (U.S.), Gaithersburg, MD (January 2023)

Tabassi, E.: Artificial Intelligence Risk Management Framework (AI RMF 1.0). Technical Report NIST AI 100-1, National Institute of Standards and Technol- ogy (U.S.), Gaithersburg, MD (January 2023). https://doi.org/10.6028/NIST.AI. 100-1 . http://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf Accessed 2025- 03-21

work page doi:10.6028/nist.ai 2023
[8]

https://eur-lex.europa.eu/legal-content/ EN/TXT/?uri=CELEX:32024R1689

European Commission: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). https://eur-lex.europa.eu/legal-content/ EN/TXT/?uri=CELEX:32024R1689. Official Journal of the European Union, L 1689/1, 13 June 2024 (2024) 24

2024
[9]

https://www.cac.gov.cn/ 2023-07/13/c 1690898327029107.htm

Cyberspace Administration of China (CAC): Interim Measures for the Man- agement of Generative Artificial Intelligence Services. https://www.cac.gov.cn/ 2023-07/13/c 1690898327029107.htm. Effective from August 15, 2023 (2023)

2023
[10]

In: 2023 ACM Confer- ence on Fairness Accountability and Transparency, pp

Golpayegani, D., Pandit, H.J., Lewis, D.: To Be High-Risk, or Not To Be—Semantic Specifications and Implications of the AI Act’s High-Risk AI Applications and Harmonised Standards. In: 2023 ACM Confer- ence on Fairness Accountability and Transparency, pp. 905–915. ACM, Chicago IL USA (2023). https://doi.org/10.1145/3593013.3594050 . https://dl.acm.org/doi...

work page doi:10.1145/3593013.3594050 2023
[11]

Yu, Y., Liu, Y., Zhang, J., Huang, Y., Wang, Y.: Understanding Generative AI Risks for Youth: A Taxonomy Based on Empirical Data. arXiv. arXiv:2502.16383 [cs] (2025). https://doi.org/10.48550/arXiv.2502.16383 . http://arxiv.org/abs/ 2502.16383 Accessed 2025-07-25

work page doi:10.48550/arxiv.2502.16383 2025
[12]

In: Proceedings of the 2025 Con- ference on Empirical Methods in Natural Language Processing (EMNLP), pp

Yong, Z.X., Ermis, B.,et al.: The state of multilingual llm safety research: From measuring the language gap to mitigating it. In: Proceedings of the 2025 Con- ference on Empirical Methods in Natural Language Processing (EMNLP), pp. 15845–15860 (2025)

2025
[13]

Computer57(11), 71–81 (2024) https://doi.org/10.1109/mc.2024.3432492

De Miguel Vel´ azquez, J., ˇS´ cepanovi´ c, S., Gvirtz, A., Quercia, D.: Decoding Real-World Artificial Intelligence Incidents. Computer57(11), 71–81 (2024) https://doi.org/10.1109/mc.2024.3432492 . Publisher: Institute of Electrical and Electronics Engineers (IEEE). Accessed 2025-07-25

work page doi:10.1109/mc.2024.3432492 2024
[14]

In: arXiv Preprint arXiv:2407.17436 (2024)

Zeng, Y.,et al.: Air-bench 2024: A safety benchmark based on risk categories from regulations and policies. In: arXiv Preprint arXiv:2407.17436 (2024)

arXiv 2024
[15]

AGI - Artificial General Intelligence - Robotics - Safety & Alignment1(1) (2024) https://doi.org/10.70777/agi.v1i1.10881

Slattery, P., Saeri, A.K., Grundy, E.A.C., Graham, J., Noetel, M., Uuk, R., Dao, J., Pour, S., Casper, S., Thompson, N.: The AI Risk Repository: A Comprehensive Meta-Review, Database, and Taxonomy of Risks From Artificial Intelligence. AGI - Artificial General Intelligence - Robotics - Safety & Alignment1(1) (2024) https://doi.org/10.70777/agi.v1i1.10881 ...

work page doi:10.70777/agi.v1i1.10881 2024
[16]

MIT AI Risk Initiative (2025)

Mylius, S., Slattery, P., et al.: The ai risk repository: December 2025 update on mapping the ai governance landscape. MIT AI Risk Initiative (2025)

2025
[17]

McGregor, S.: Preventing Repeated Real World AI Failures by Cataloging Inci- dents: The AI Incident Database. arXiv. arXiv:2011.08512 [cs] (2020). https:// doi.org/10.48550/arXiv.2011.08512 . http://arxiv.org/abs/2011.08512 Accessed 2025-07-25

work page doi:10.48550/arxiv.2011.08512 2011
[18]

https://www.aiaaic.org/

Pownall, C., Associates, C..: AIAAIC: Artificial Intelligence, Algorithms, and Automation Incident & Controversy Repository. https://www.aiaaic.org/. Accessed: 2025-08-01 (2023) 25

2025
[19]

Knight, S.: Learning about AI ethics from cases: A scoping review of AI incident repositories and cases (2025)

2025
[20]

, author Cui, I

Turri, V., Dzombak, R.: Why We Need to Know More: Exploring the State of AI Incident Documentation Practices. In: Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, And Society, pp. 576–583. ACM, Montr´ eal QC Canada (2023). https://doi.org/10.1145/3600211.3604700 . https://dl.acm.org/doi/10.1145/3600211.3604700Accessed 2025-07-24

work page doi:10.1145/3600211.3604700 2023
[21]

Law, Technology and Humans5(1), 133–152 (2023) https://doi.org/ 10.5204/lthj.2682

Lupo, G.: Risky Artificial Intelligence: The Role of Incidents in the Path to AI Regulation. Law, Technology and Humans5(1), 133–152 (2023) https://doi.org/ 10.5204/lthj.2682 . Publisher: Queensland University of Technology. Accessed 2025-07-25

work page doi:10.5204/lthj.2682 2023
[22]

AI and Ethics6(2) (2026) https://doi.org/10.1007/s43681-025-00700-x

Gipiˇ skis, A., et al.: Uncovering ai’s hidden risks: an empirical analysis of health- related ai incidents and their ethical implications. AI and Ethics6(2) (2026) https://doi.org/10.1007/s43681-025-00700-x

work page doi:10.1007/s43681-025-00700-x 2026
[23]

Paeth, K., Atherton, D., Pittaras, N., Frase, H., McGregor, S.: Lessons for Edi- tors of AI Incidents from the AI Incident Database. arXiv. arXiv:2409.16425 [cs] (2024). https://doi.org/10.48550/arXiv.2409.16425 . http://arxiv.org/abs/2409. 16425 Accessed 2025-07-25

work page doi:10.48550/arxiv.2409.16425 2024
[24]

Rao, P., ˇS´ cepanovi´ c, S., et al.: RiskRAG: A Data-Driven Solution for Improved AI Model Risk Reporting (2025)

2025
[25]

Technical Report 34, OECD (February 2025)

OECD: Towards a common reporting framework for ai incidents. Technical Report 34, OECD (February 2025). Accessed: 2026-03-19. https://www.oecd. org/en/publications/towards-a-common-reporting-framework-for-ai-incidents f326d4ac-en.html

2025
[26]

https: //arxiv.org/abs/2412.14855

Bieringer, L., Paeth, K., St¨ angler, J., Wespi, A., Alahi, A., Grosse, K.: Position: A Taxonomy for Reporting and Describing AI Security Incidents (2025). https: //arxiv.org/abs/2412.14855

Pith/arXiv arXiv 2025
[27]

https: //arxiv.org/abs/2507.23669

Russo, D., Orlando, G.M., La Gatta, V., Moscato, V.: Automating AI Failure Tracking: Semantic Association of Reports in AI Incident Database (2025). https: //arxiv.org/abs/2507.23669

arXiv 2025
[28]

https://arxiv.org/abs/2505.04291

Richards, I., Benn, C., Zilka, M.: From Incidents to Insights: Patterns of Responsibility following AI Harms (2025). https://arxiv.org/abs/2505.04291

arXiv 2025
[29]

Popchanovska, E., et al.: When AI Fails, What Works? A Data-Driven Taxonomy of Real-World AI Risk Mitigation Strategies (2026)

2026
[30]

In: 17th ACM/SPEC International 26 Conference on Performance Engineering (ICPE 2026) (2026)

Chu, X., Ilager, S.,et al.: Leveraging llms for structured information extrac- tion and analysis from cloud incident reports. In: 17th ACM/SPEC International 26 Conference on Performance Engineering (ICPE 2026) (2026). https://doi.org/10. 1145/3777911.3801103

arXiv 2026
[31]

https://data

Common Crawl: Common Crawl News Dataset Index. https://data. commoncrawl.org/crawl-data/CC-NEWS/index.html. Accessed: 2025-08-01 (2016)

2025
[32]

https://arxiv.org/abs/ 2407.13773

He, C., Li, W., Jin, Z., Xu, C., Wang, B., Lin, D.: OpenDataLab: Empowering General Artificial Intelligence with Open Datasets (2024). https://arxiv.org/abs/ 2407.13773

arXiv 2024
[33]

https://github

SnailDev: Hot Hub repositories for Chinese online trending topics. https://github. com/snaildev. Including Weibo Hot Hub, Douyin Hot Hub, Zhihu Hot Hub, Toutiao Hot Hub, and V2EX Hot Hub. Accessed: 2025-08-01 (2021) 27

2025

[1] [1]

Technical report, Depart- ment for Science, Innovation and Technology (UK) (2026)

Bengio, Y., et al.: International ai safety report 2026. Technical report, Depart- ment for Science, Innovation and Technology (UK) (2026). A global synthesis of scientific evidence on GPAI risks. https://arxiv.org/abs/2602.21012

arXiv 2026

[2] [2]

Managing extreme AI risks amid rapid progress

Bengio, Y., Hinton, G., Yao, A., Song, D., Abbeel, P., Darrell, T., Harari, Y.N., Zhang, Y.-Q., Xue, L., Shalev-Shwartz, S., Hadfield, G., Clune, J., Maharaj, T., Hutter, F., Baydin, A.G., McIlraith, S., Gao, Q., Acharya, A., Krueger, D., Dragan, A., Torr, P., Russell, S., Kahneman, D., Brauner, J., Mindermann, S.: Managing extreme AI risks amid rapid pro...

work page doi:10.1126/science.adn0117 2024

[3] [3]

Winfield, A.F.T., Jirotka, M.: Ethical governance is essential to building trust in robotics and artificial intelligence systems. Philosophical Trans- actions of the Royal Society A: Mathematical, Physical and Engineering Sciences376(2133), 20180085 (2018) https://doi.org/10.1098/rsta.2018.0085 https://royalsocietypublishing.org/doi/pdf/10.1098/rsta.2018.0085

work page doi:10.1098/rsta.2018.0085 2018

[4] [4]

Big Data & Society , author =

Mittelstadt, B.D., Allo, P., Taddeo, M., Wachter, S., Floridi, L.: The ethics of algorithms: Mapping the debate. Big Data & Society3(2), 2053951716679679 (2016) https://doi.org/10.1177/2053951716679679 https://doi.org/10.1177/2053951716679679

work page doi:10.1177/2053951716679679 2016

[5] [5]

TIME Magazine (2025)

Atherton, K.D.: Reports of ai-related incidents rose 50% year-over-year: The new kinds of ai harm emerging. TIME Magazine (2025)

2025

[6] [6]

Staufer, L., Casper, S., et al.: The 2025 AI Agent Index: Documenting Technical and Safety Features of Deployed Agentic AI Systems (2026)

2025

[7] [7]

Technical Report NIST AI 100-1, National Institute of Standards and Technol- ogy (U.S.), Gaithersburg, MD (January 2023)

Tabassi, E.: Artificial Intelligence Risk Management Framework (AI RMF 1.0). Technical Report NIST AI 100-1, National Institute of Standards and Technol- ogy (U.S.), Gaithersburg, MD (January 2023). https://doi.org/10.6028/NIST.AI. 100-1 . http://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf Accessed 2025- 03-21

work page doi:10.6028/nist.ai 2023

[8] [8]

https://eur-lex.europa.eu/legal-content/ EN/TXT/?uri=CELEX:32024R1689

European Commission: Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). https://eur-lex.europa.eu/legal-content/ EN/TXT/?uri=CELEX:32024R1689. Official Journal of the European Union, L 1689/1, 13 June 2024 (2024) 24

2024

[9] [9]

https://www.cac.gov.cn/ 2023-07/13/c 1690898327029107.htm

Cyberspace Administration of China (CAC): Interim Measures for the Man- agement of Generative Artificial Intelligence Services. https://www.cac.gov.cn/ 2023-07/13/c 1690898327029107.htm. Effective from August 15, 2023 (2023)

2023

[10] [10]

In: 2023 ACM Confer- ence on Fairness Accountability and Transparency, pp

Golpayegani, D., Pandit, H.J., Lewis, D.: To Be High-Risk, or Not To Be—Semantic Specifications and Implications of the AI Act’s High-Risk AI Applications and Harmonised Standards. In: 2023 ACM Confer- ence on Fairness Accountability and Transparency, pp. 905–915. ACM, Chicago IL USA (2023). https://doi.org/10.1145/3593013.3594050 . https://dl.acm.org/doi...

work page doi:10.1145/3593013.3594050 2023

[11] [11]

Yu, Y., Liu, Y., Zhang, J., Huang, Y., Wang, Y.: Understanding Generative AI Risks for Youth: A Taxonomy Based on Empirical Data. arXiv. arXiv:2502.16383 [cs] (2025). https://doi.org/10.48550/arXiv.2502.16383 . http://arxiv.org/abs/ 2502.16383 Accessed 2025-07-25

work page doi:10.48550/arxiv.2502.16383 2025

[12] [12]

In: Proceedings of the 2025 Con- ference on Empirical Methods in Natural Language Processing (EMNLP), pp

Yong, Z.X., Ermis, B.,et al.: The state of multilingual llm safety research: From measuring the language gap to mitigating it. In: Proceedings of the 2025 Con- ference on Empirical Methods in Natural Language Processing (EMNLP), pp. 15845–15860 (2025)

2025

[13] [13]

Computer57(11), 71–81 (2024) https://doi.org/10.1109/mc.2024.3432492

De Miguel Vel´ azquez, J., ˇS´ cepanovi´ c, S., Gvirtz, A., Quercia, D.: Decoding Real-World Artificial Intelligence Incidents. Computer57(11), 71–81 (2024) https://doi.org/10.1109/mc.2024.3432492 . Publisher: Institute of Electrical and Electronics Engineers (IEEE). Accessed 2025-07-25

work page doi:10.1109/mc.2024.3432492 2024

[14] [14]

In: arXiv Preprint arXiv:2407.17436 (2024)

Zeng, Y.,et al.: Air-bench 2024: A safety benchmark based on risk categories from regulations and policies. In: arXiv Preprint arXiv:2407.17436 (2024)

arXiv 2024

[15] [15]

AGI - Artificial General Intelligence - Robotics - Safety & Alignment1(1) (2024) https://doi.org/10.70777/agi.v1i1.10881

Slattery, P., Saeri, A.K., Grundy, E.A.C., Graham, J., Noetel, M., Uuk, R., Dao, J., Pour, S., Casper, S., Thompson, N.: The AI Risk Repository: A Comprehensive Meta-Review, Database, and Taxonomy of Risks From Artificial Intelligence. AGI - Artificial General Intelligence - Robotics - Safety & Alignment1(1) (2024) https://doi.org/10.70777/agi.v1i1.10881 ...

work page doi:10.70777/agi.v1i1.10881 2024

[16] [16]

MIT AI Risk Initiative (2025)

Mylius, S., Slattery, P., et al.: The ai risk repository: December 2025 update on mapping the ai governance landscape. MIT AI Risk Initiative (2025)

2025

[17] [17]

McGregor, S.: Preventing Repeated Real World AI Failures by Cataloging Inci- dents: The AI Incident Database. arXiv. arXiv:2011.08512 [cs] (2020). https:// doi.org/10.48550/arXiv.2011.08512 . http://arxiv.org/abs/2011.08512 Accessed 2025-07-25

work page doi:10.48550/arxiv.2011.08512 2011

[18] [18]

https://www.aiaaic.org/

Pownall, C., Associates, C..: AIAAIC: Artificial Intelligence, Algorithms, and Automation Incident & Controversy Repository. https://www.aiaaic.org/. Accessed: 2025-08-01 (2023) 25

2025

[19] [19]

Knight, S.: Learning about AI ethics from cases: A scoping review of AI incident repositories and cases (2025)

2025

[20] [20]

, author Cui, I

Turri, V., Dzombak, R.: Why We Need to Know More: Exploring the State of AI Incident Documentation Practices. In: Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, And Society, pp. 576–583. ACM, Montr´ eal QC Canada (2023). https://doi.org/10.1145/3600211.3604700 . https://dl.acm.org/doi/10.1145/3600211.3604700Accessed 2025-07-24

work page doi:10.1145/3600211.3604700 2023

[21] [21]

Law, Technology and Humans5(1), 133–152 (2023) https://doi.org/ 10.5204/lthj.2682

Lupo, G.: Risky Artificial Intelligence: The Role of Incidents in the Path to AI Regulation. Law, Technology and Humans5(1), 133–152 (2023) https://doi.org/ 10.5204/lthj.2682 . Publisher: Queensland University of Technology. Accessed 2025-07-25

work page doi:10.5204/lthj.2682 2023

[22] [22]

AI and Ethics6(2) (2026) https://doi.org/10.1007/s43681-025-00700-x

Gipiˇ skis, A., et al.: Uncovering ai’s hidden risks: an empirical analysis of health- related ai incidents and their ethical implications. AI and Ethics6(2) (2026) https://doi.org/10.1007/s43681-025-00700-x

work page doi:10.1007/s43681-025-00700-x 2026

[23] [23]

Paeth, K., Atherton, D., Pittaras, N., Frase, H., McGregor, S.: Lessons for Edi- tors of AI Incidents from the AI Incident Database. arXiv. arXiv:2409.16425 [cs] (2024). https://doi.org/10.48550/arXiv.2409.16425 . http://arxiv.org/abs/2409. 16425 Accessed 2025-07-25

work page doi:10.48550/arxiv.2409.16425 2024

[24] [24]

Rao, P., ˇS´ cepanovi´ c, S., et al.: RiskRAG: A Data-Driven Solution for Improved AI Model Risk Reporting (2025)

2025

[25] [25]

Technical Report 34, OECD (February 2025)

OECD: Towards a common reporting framework for ai incidents. Technical Report 34, OECD (February 2025). Accessed: 2026-03-19. https://www.oecd. org/en/publications/towards-a-common-reporting-framework-for-ai-incidents f326d4ac-en.html

2025

[26] [26]

https: //arxiv.org/abs/2412.14855

Bieringer, L., Paeth, K., St¨ angler, J., Wespi, A., Alahi, A., Grosse, K.: Position: A Taxonomy for Reporting and Describing AI Security Incidents (2025). https: //arxiv.org/abs/2412.14855

Pith/arXiv arXiv 2025

[27] [27]

https: //arxiv.org/abs/2507.23669

Russo, D., Orlando, G.M., La Gatta, V., Moscato, V.: Automating AI Failure Tracking: Semantic Association of Reports in AI Incident Database (2025). https: //arxiv.org/abs/2507.23669

arXiv 2025

[28] [28]

https://arxiv.org/abs/2505.04291

Richards, I., Benn, C., Zilka, M.: From Incidents to Insights: Patterns of Responsibility following AI Harms (2025). https://arxiv.org/abs/2505.04291

arXiv 2025

[29] [29]

Popchanovska, E., et al.: When AI Fails, What Works? A Data-Driven Taxonomy of Real-World AI Risk Mitigation Strategies (2026)

2026

[30] [30]

In: 17th ACM/SPEC International 26 Conference on Performance Engineering (ICPE 2026) (2026)

Chu, X., Ilager, S.,et al.: Leveraging llms for structured information extrac- tion and analysis from cloud incident reports. In: 17th ACM/SPEC International 26 Conference on Performance Engineering (ICPE 2026) (2026). https://doi.org/10. 1145/3777911.3801103

arXiv 2026

[31] [31]

https://data

Common Crawl: Common Crawl News Dataset Index. https://data. commoncrawl.org/crawl-data/CC-NEWS/index.html. Accessed: 2025-08-01 (2016)

2025

[32] [32]

https://arxiv.org/abs/ 2407.13773

He, C., Li, W., Jin, Z., Xu, C., Wang, B., Lin, D.: OpenDataLab: Empowering General Artificial Intelligence with Open Datasets (2024). https://arxiv.org/abs/ 2407.13773

arXiv 2024

[33] [33]

https://github

SnailDev: Hot Hub repositories for Chinese online trending topics. https://github. com/snaildev. Including Weibo Hot Hub, Douyin Hot Hub, Zhihu Hot Hub, Toutiao Hot Hub, and V2EX Hot Hub. Accessed: 2025-08-01 (2021) 27

2025