The Risks of Machine Learning Systems
A structured ML risk taxonomy covering safety, privacy, discrimination, and security could inform APS risk assessment templates and AI governance frameworks.
Key points
- MIT AI Risk Repository spotlights the 2022 MLSR framework, categorising ML risks into first-order and second-order types.
- The framework offers a structured taxonomy integrating impact assessments, incident reports, and ML literature - useful for risk assessment design.
- This is a 2022 academic paper being surfaced via a blog digest; it is reference material rather than new guidance.
Implications for Australian agencies
- Consider AI governance and risk teams could assess whether the MLSR first/second-order risk structure complements or gaps existing agency risk assessment templates for AI systems.
- Monitor Practitioners tracking the MIT AI Risk Repository may want to watch the full repository for additional frameworks with direct applicability to APS AI governance contexts.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
"The Risks of Machine Learning Systems"
Source: MIT AI Risk Repository – Blog
Published: 23 April 2025
URL: https://airisk.mit.edu/blog/the-risks-of-machine-learning-systems
The MIT AI Risk Repository blog spotlights the Machine Learning System Risk (MLSR) framework by Tan, Taeihagh, and Baxter (2022). The framework distinguishes between first-order risks arising from system design and implementation choices (including algorithm robustness, data quality, and emergent behaviour) and second-order risks that emerge when those systems interact with the real world (including safety, privacy, discrimination, and security harms). Its taxonomy draws on algorithmic impact assessments, incident reports, software risk literature, and professional experience. For APS practitioners, the framework offers a structured starting point for conducting or reviewing ML system risk assessments, though it predates several more recent Australian and international governance developments.
Implications for Australian agencies:
- [Consider] AI governance and risk teams could assess whether the MLSR first/second-order risk structure complements or gaps existing agency risk assessment templates for AI systems.
- [Monitor] Practitioners tracking the MIT AI Risk Repository may want to watch the full repository for additional frameworks with direct applicability to APS AI governance contexts.
Retrieved from SIMS, 18 July 2026.