Mapping the AI Governance Landscape: Pilot Test and Update
A systematic map of what AI governance frameworks actually cover—and what they neglect—gives APS policy teams an evidence base for identifying gaps in Australia's own framework coverage.
Key points
- MIT AI Risk Repository used LLMs to classify 950+ AI governance documents across risk, mitigation, and sector taxonomies.
- Governance failure, security vulnerabilities, and transparency were the most-covered risk domains; AI welfare and multi-agent risks were least covered.
- US-heavy dataset limits global generalisability; Australian documents are unlikely to be well-represented in current outputs.
Implications for Australian agencies
- Monitor Policy teams working on AI risk governance frameworks may want to monitor the forthcoming MIT database and preprint to benchmark Australian framework coverage against the global landscape.
- Consider APS agencies involved in AI governance design could consider whether the MIT AI Risk Taxonomy and mitigation taxonomy provide a useful structured lens for auditing coverage gaps in existing Australian guidance documents.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
"Mapping the AI Governance Landscape: Pilot Test and Update"
Source: MIT AI Risk Repository – Blog
Published: 15 October 2025
URL: https://airisk.mit.edu/blog/mapping-the-ai-governance-landscape-pilot-test-and-update
The MIT AI Risk Repository has piloted an LLM-assisted pipeline to classify over 950 AI governance documents from CSET's AGORA archive using the MIT AI Risk Taxonomy and related frameworks. The study found that top LLMs (Claude Opus 4.1, GPT-5, Claude Sonnet 4.5) achieved inter-rater agreement with human reviewers comparable to or exceeding agreement between two independent human coders. Preliminary findings show that governance documents cluster heavily around governance failure, security vulnerabilities, and transparency risks, while AI welfare, multi-agent risks, and economic devaluation of human effort receive scant coverage. The dataset is US-centric and English-only, which limits applicability to global or Australian-specific governance mapping, but the planned outputs—including visualisations, a searchable database, and a preprint—could be a useful reference tool for APS teams assessing coverage gaps in Australian AI frameworks.
Implications for Australian agencies:
- [Monitor] Policy teams working on AI risk governance frameworks may want to monitor the forthcoming MIT database and preprint to benchmark Australian framework coverage against the global landscape.
- [Consider] APS agencies involved in AI governance design could consider whether the MIT AI Risk Taxonomy and mitigation taxonomy provide a useful structured lens for auditing coverage gaps in existing Australian guidance documents.
Retrieved from SIMS, 18 July 2026.