Sources of Risk of AI Systems
A structured AI risk taxonomy may inform how APS agencies categorise and assess risks in their own AI governance frameworks.
Key points
- MIT AI Risk Repository spotlights a 2022 taxonomy classifying AI risk sources into ethical and reliability/robustness clusters.
- The framework integrates AI risk sources into formal risk assessment processes, distinguishing ML systems from classical software.
- This is a 2022 academic paper surfaced via a blog spotlight - not new guidance or a regulatory development.
Implications for Australian agencies
- Consider Agencies developing or refining AI risk assessment frameworks could consider whether this taxonomy's ethical/reliability split maps usefully onto their existing risk categorisation approaches.
- Monitor Policy teams may want to monitor the MIT AI Risk Repository more broadly as a curated source of risk frameworks that could inform Australian AI governance thinking.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
"Sources of Risk of AI Systems"
Source: MIT AI Risk Repository – Blog
Published: 4 March 2025
URL: https://airisk.mit.edu/blog/sources-of-risk-of-ai-systems
The MIT AI Risk Repository's blog has spotlighted a 2022 peer-reviewed paper by Steimers and Schneider that proposes a taxonomy of AI-specific risk sources. The taxonomy divides risks into ethical dimensions (fairness, privacy, automation and control) and reliability/robustness dimensions (task complexity, transparency, explainability, security, hardware, and technological maturity). The paper also outlines a risk management process integrating these sources into system-level risk assessment, and distinguishes AI systems built on modern machine learning from classical software. The blog entry is a summary spotlight rather than new research or regulatory output.
Implications for Australian agencies:
- [Consider] Agencies developing or refining AI risk assessment frameworks could consider whether this taxonomy's ethical/reliability split maps usefully onto their existing risk categorisation approaches.
- [Monitor] Policy teams may want to monitor the MIT AI Risk Repository more broadly as a curated source of risk frameworks that could inform Australian AI governance thinking.
Retrieved from SIMS, 18 July 2026.