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

2026-06-07 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cybersecurity & Data Privacy · Depth: Advanced, quick

Summary

RiskNet is a new, large-scale dataset designed to track and analyze real-world AI risk incidents, addressing the current scarcity of empirical resources. Constructed from hundreds of millions of multilingual news sources, RiskNet employs a structured pipeline for identifying AI risk news, screening event-level reports, aligning incidents, and performing multi-dimensional classification. This resource organizes dispersed news reports into incident-centered records, offering benchmark datasets for event classification, incident alignment, and incident-level risk labeling. The current release includes aligned incident clusters and annotated benchmark subsets, accessible via an online platform. RiskNet aims to support research in AI safety, governance, risk analysis, and benchmarking, providing a structured resource to bridge the gap between high-level AI governance principles and documented real-world harms.

Key takeaway

For AI scientists and governance researchers developing or evaluating responsible AI frameworks, RiskNet provides an essential empirical foundation. You should integrate this large-scale dataset into your risk analysis and benchmarking efforts to move beyond high-level principles. Utilize its incident clusters and annotated subsets to validate models and inform policy, ensuring your work reflects real-world AI harms.

Key insights

A large-scale, structured dataset of real-world AI risk incidents is crucial for empirical AI safety and governance research.

Principles

• AI risk tracking needs empirical data.
• News sources offer rich incident data.
• Structured data bridges governance gaps.

Method

RiskNet uses a pipeline: AI risk news identification, event-level report screening, incident alignment, and multi-dimensional incident classification from multilingual news.

In practice

• Use RiskNet for AI safety research.
• Explore incidents via online platform.
• Benchmark AI risk classification models.

Topics

• AI Risk Incidents
• AI Safety
• AI Governance
• RiskNet Dataset
• Machine Learning
• Dataset Annotation

Best for: Research Scientist, AI Scientist, AI Ethicist, Policy Maker

Related on AIssential

• Towards AI epidemiology: a measurement standardisation framework for prospective risk detection — A framework enables "AI epidemiology" by standardizing expert-AI interactions for prospective risk detection without model internals.
• The inaugural Redwood Research podcast — AI safety requires pragmatic strategies, acknowledging both technical challenges and organizational realities, to manage escalating risks.
• AI Safety: How do we reduce harm and navigate bias in AI? — AI safety prioritizes harm reduction, especially for vulnerable groups, by identifying and mitigating inherent biases.
• ChinAI #343: AI Safety/Security Governance Research Report (CAICT 2025) — Chinese AI governance research actively integrates global findings on safety, security, and emerging risks.
• OpenAI’s Frontier Governance Framework — OpenAI's Frontier Governance Framework aligns advanced AI safety practices with emerging global legal requirements.
• ‘In the end, you feel blank’: India’s female workers watching hours of abusive content to train AI — Content moderation for AI training inflicts severe psychological trauma on workers, particularly women in rural India.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.