Risks from Malfunctions

Classification

Domain

7. AI system safety, failures, & limitations

Subdomain

7.0

Causal entity

Other

Intent

Unintentional

Timing

Other

How other frameworks describe this risk

• Operational harms (financial markets)

  Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems (Gipiškis2024)

• Fine-tuning related (Degrading safety training due to benign fine-tuning)

  Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems (Gipiškis2024)

• Safety & Trustworthiness

  Cataloguing LLM Evaluations (InfoComm2023)

• Extreme Risks

  Cataloguing LLM Evaluations (InfoComm2023)

• Limitations of Human Feedback

  AI Alignment: A Comprehensive Survey (Ji2023)

• Inappropriate data splitting

  AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks (Schnitzer2024)

• Uncertainty concerns

  AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks (Schnitzer2024)

• Choice of untrustworthy data source

  AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks (Schnitzer2024)

Other entries from Bengio2024

• Malicious Use Risks Risk Category · 4.0
• Harm to individuals through fake content Risk Sub-Category · 4.3
• Harm to individuals through fake content Additional evidence
• Disinformation and manipulation of public opinion Risk Sub-Category · 4.1
• Cyber offence Risk Sub-Category · 4.2
• Cyber offence Additional evidence
• Cyber offence Additional evidence
• Dual use science risks Risk Sub-Category · 4.2
• Risks from product functionality issues Risk Sub-Category · 5.1
• Risks from product functionality issues Additional evidence
• Risks from product functionality issues Additional evidence
• Risks from product functionality issues Additional evidence