Loading
10 items curated by AIRiskAware · Published 28 May 2026
When this issue went out, the Digital Omnibus on AI had been provisionally agreed on 7 May 2026 and adoption was still pending. Updated since publication: it was adopted on 8 July 2026 as Regulation (EU) 2026/1744 and has been in force since 27 July 2026. This link is the consolidated text of the AI Act as amended by it, which is the version to work from rather than the original 2024 text.
AIRiskAware take
Check your organisation's AI Act implementation against this version, not the 2024 text. The changes to high-risk AI obligations under Annex III are material for healthcare and financial services deployers.
Phase 2 of Australia's Privacy Act reforms commenced in 2026, including new requirements around automated decision-making disclosure and enhanced enforcement powers for the OAIC.
AIRiskAware take
If you're an APP entity and haven't updated your privacy policy to address AI-assisted decisions, the window for doing this voluntarily is now closing.
The FTC continued its enforcement posture toward AI-washing, deceptive AI claims, and consumer harm from algorithmic systems. The FTC blog is the fastest way to track US AI enforcement without legal subscriptions.
AIRiskAware take
Three consistent themes: (1) marketing claims about AI capabilities need substantiation; (2) "AI explains" is not a valid response to an adverse action, actual explanations are required; (3) children's data is a near-zero-tolerance category.
An ECB Research Bulletin running simulation experiments on two AI architectures to see which destabilises a deposit market. Directly relevant to financial stability risk frameworks.
AIRiskAware take
The result cuts against the intuitive story. Q-learning agents coordinated tightly and were prone to extreme bank-run-like dynamics; large language models were LESS prone to runs but behaved heterogeneously and unpredictably. The risk is the architecture, not the label AI, so firms should ask which class of model sits behind an automated decision rather than whether one does.
NIST continues to publish profiles and supplementary guidance against the AI RMF 1.0 framework. The playbook is where the operational implementation detail lives.
AIRiskAware take
If your governance team cites "NIST AI RMF compliance" without referencing the playbook, they're working at the wrong level of abstraction.
The EU Platform Work Directive includes specific requirements on algorithmic management for gig economy platforms: transparency, contestability, and human review. In force from 2026.
AIRiskAware take
Often missed by AI governance teams who focus on the AI Act. If your organisation uses AI to manage contractors or gig workers, this Directive applies independently of the AI Act and has its own obligations.
Singapore continues to update the Model AI Governance Framework with sector-specific guidance. The PDPC remains one of the most practically useful regulators for governance teams operating in APAC.
AIRiskAware take
The Singapore framework is unusually implementation-focused compared to EU or US guidance. It's worth reading even if you're not operating in Singapore, the operational checklists translate across jurisdictions.
The annual AI Index is the most comprehensive quantitative assessment of AI capabilities, adoption, and policy. The 2026 edition covers regulatory developments across 50+ countries.
AIRiskAware take
Chapter 4 (policy and governance) is where to start for practitioners. The data on AI patent filings by jurisdiction, enforcement action counts, and regulatory publication rates is uniquely useful for benchmarking.
The OECD AI Observatory tracks regulatory developments across member countries. The AI Principles were updated in 2024 and the implementing guidance continues to evolve.
AIRiskAware take
Underused by practitioners. The comparative policy tracker is genuinely useful for multi-jurisdictional compliance work.
Australia's national science agency continues to publish practical AI governance research, including work on AI evaluation, bias testing, and governance frameworks for Australian contexts.
AIRiskAware take
The CSIRO research on evaluating AI in high-stakes Australian government contexts is among the most practically grounded work in the APAC region.
Keep reading
New issues are published here each month. Browse the full archive, or explore our in-depth analysis, every piece linked to primary regulatory sources.