NeuralStackAI Law Radar has issued a broad set of corrections and additions to its AI regulation tracker, tightening several entries and changing how some of the underlying instruments are described. The latest revision log says the tracker was last verified on 12 August 2026 and now reflects material updates across copyright, privacy, transparency and AI supervision topics.
One of the main corrections concerns Saudi Arabia’s AI training exemption. The tracker says a previously cited Umm Al-Qura URL resolved to an earlier Cabinet Resolution rather than the enacted Royal Decree M/169 text, so the source link was corrected to the gazette page publishing the Copyright Law itself. It also says the Art.60 Implementing Regulation, described as the Executive Regulation of the Copyright System, has now been published in Umm Al-Qura ahead of the law’s entry into force.
The Vietnam entry was also revised after the tracker found that supplementary guidance from the Ministry of Science and Technology had already been published. According to the update, the ministry posted explainer material on 3 July 2026 and a further notice on 8 July 2026, covering the six sectors designated under Decision 33/2026/QD-TTg and the two transition deadlines. The substance of the obligation is unchanged, but the earlier statement that no supplementary guidance existed was removed.
Canada’s status note was updated to reflect a different federal bill rather than a revived comprehensive AI act. AI Law Radar says AIDA under Bill C-27 remains dead on prorogation, but Bill C-36, the Protecting Privacy and Consumer Data Act, had first reading on 15 June 2026 and was at second reading in the House as of 11 August 2026. The tracker describes it as a privacy-law vehicle with AI transparency provisions, including disclosure for automated or algorithmic decision tools.
Singapore saw a more significant reclassification. The tracker says SAFR, or Safeguards for Agentic Finance at Runtime, is not MAS’s binding supervisory framework for agentic AI but an industry-led voluntary information paper published on 3 July 2026. The binding track remains MAS’s proposed AI Risk Management Guidelines, which are still in consultation and do not yet have a confirmed finalization date.
The tracker also expanded its copyright-and-training coverage with entries for Japan, Singapore and the EU. It says Japan’s Copyright Act Art.30-4 is the broadest of the three, Singapore’s Copyright Act 2021 s.244 allows copying for computational data analysis under conditions tied to lawful access, and the EU’s DSM Directive Art. 4 allows commercial mining of lawfully accessible works unless the rightholder has reserved the use by machine-readable means. The EU AI Act cross-references that reservation rule for general-purpose AI models, which is why the two EU entries are linked rather than duplicated.
Why this matters
For developers and compliance teams, the biggest takeaway is that AI governance is splitting into distinct legal layers rather than one unified regime. The tracker’s corrections show how easily a voluntary industry paper, a privacy bill with AI clauses, and a binding supervisory framework can be conflated if the source material is not checked carefully.
The changes also underscore that training-data rights are now being handled differently across jurisdictions. Japan, Singapore and the EU each permit some form of text-and-data mining or computational analysis, but they do so with different limits on access, reservation, and contractual override, which complicates cross-border model training and policy design.
On the transparency side, the tracker’s revised entries reinforce that legal obligations are arriving in fragments. Some rules focus on disclosure to end users, others on content marking, and others on privacy notices or supervisory expectations, which means vendors may need to align one product with several regimes at once.
The UK check in the update is also notable because it found no new entry despite review of recent legislation. AI Law Radar says the Data (Use and Access) Act 2025 places reporting duties on the Secretary of State rather than on AI developers, while the older research exception remains limited to non-commercial use.
For now, the tracker’s broader message is that the regulatory map keeps changing by jurisdiction and by legal category. The latest round of edits does not just add new entries; it also narrows them, drawing a sharper line between binding law, policy guidance and industry-led materials.