NeuralStackAI Law Radar’s latest update log sketches a fast‑moving regulatory landscape where lawmakers are fine‑tuning how artificial intelligence can be built and used, often by closing specific gaps rather than passing broad, standalone AI acts. The register, which tracks material changes to AI‑related obligations, confirms that the United Kingdom, China and several US states have all either introduced new rules or clarified the scope and timing of existing ones. Across the entries, the emphasis falls on automated decision‑making, synthetic intimate imagery and algorithmic price coordination, signalling growing discomfort with unregulated high‑impact AI uses.
In the United Kingdom, AI Law Radar has upgraded its confidence rating on the AI (Regulation) Bill in the House of Lords, but the underlying story is one of inertia rather than progress. The site notes that the official Bills API shows the proposal still stuck at first reading, with no movement since its introduction, and points to the government’s own Blueprint for AI Regulation favouring an “AI Growth Lab” and sector‑specific sandboxes over primary legislation. That combination leads AI Law Radar to reinforce its assessment that the bill is unlikely to advance in its current form, even as it continues to track the measure as a proposed framework rather than a live obligation.
More concretely, the UK’s Data (Use and Access) Act 2025 is already reshaping automated decision‑making rules under the UK GDPR. AI Law Radar highlights new sections 22A to 22D, which took effect in early February and replace the old default prohibition on solely automated decisions with a structure built around notification duties, representation rights, human review and the ability to contest outcomes. In parallel, section 138 of the same act inserts new offences into the Sexual Offences Act 2003 that criminalise creating or even requesting a non‑consensual intimate deepfake, regardless of whether that content is ever shared, a carve‑out the tracker carefully distinguishes from newer offences aimed at tool suppliers in the Crime and Policing Act.
The handling of those deepfake provisions also illustrates AI Law Radar’s role as a meta‑editor of regulatory intelligence rather than just a collector of citations. One correction entry notes that the obligations tied to DUAA 2025 section 138 were initially scoped to “deployers” of AI, but on closer reading the law binds any person in the UK who creates or requests non‑consensual intimate synthetic imagery. The site has since broadened the scope label to “everyone” while keeping the dates and sources intact, and has similarly tightened its tagging of algorithmic price‑coordination bans so that related US state laws appear under a unified “Prohibited AI practices” theme page.
US states dominate the algorithmic pricing story. AI Law Radar’s changelog records Illinois SB 343, a bill that amends the Illinois Antitrust Act to explicitly ban algorithmic coordination of rental prices. The measure has cleared the state’s General Assembly and sits on the governor’s desk with an action deadline at the end of August, leading the tracker to classify it as proposed and without an effective date for now. In a related clean‑up, the site retags Illinois SB 343, New Jersey’s FAIR Act and Maryland’s Protection From Predatory Pricing Act under the same prohibited‑AI rubric, arguing that each one directly bans the use of algorithms for certain kinds of price setting.
The update feed also surfaces a new Chinese instrument aimed at the emerging class of AI “agents.” AI Law Radar has added the CAC, NDRC and MIIT’s joint AI Agents Implementation Opinions, issued in early May, to its coverage. While the changelog does not unpack the full contents, the decision to list the Opinions alongside hard‑law obligations suggests regulators in Beijing are now paying targeted attention to anthropomorphic or agent‑like AI systems, mirroring a global pattern in which conversational and interactive models draw bespoke scrutiny.
Beyond new entries, the tracker spends considerable effort on tightening the accuracy of its existing records, a sign of how fluid AI law can be even after statutes pass. One correction clarifies that New York’s RAISE Act on frontier AI safety was mis‑cited and in fact became law under a different bill number, though its effective date remains in the future. Another fixes the citation for Australia’s new automated decision‑making transparency duty under the Privacy Act, confirming that the obligations will sit in APP 1.7 to 1.9 with a commencement date unchanged, and several US and South Korean rows see their confidence levels raised after official briefings and press releases confirm previously uncertain effective dates and grace periods.
Why this matters
For companies building or deploying AI systems, the AI Law Radar changelog reads less like abstract legislative theatre and more like a growing checklist of very specific behaviours that are either newly criminalised or subject to heightened procedural safeguards. The UK’s shift from a blanket ban on automated decisions to detailed duties around notice, human review and contest rights underscores that regulators are moving from symbolic alarm to operational guardrails, while the deepfake offences show lawmakers are willing to reach into criminal law to address harms that may never leave a private device. On the commercial side, state‑level bans on algorithmic rent coordination and similar pricing practices make clear that antitrust enforcers will not treat algorithmic collusion as a loophole, raising the stakes for landlords, platforms and data‑analytics firms that lean on optimisation software.
Looking ahead, the pattern in AI Law Radar’s updates suggests that AI regulation will continue to emerge as a patchwork of sector‑specific duties, targeted criminal offences and competition‑law tweaks rather than singular omnibus statutes, particularly in jurisdictions like the UK where the political appetite for a cross‑cutting AI act appears limited. At the same time, the introduction of Chinese AI Agents Opinions and the careful attention to grace periods and effective dates in places like South Korea and various US states hint at a global convergence on phased, risk‑tiered approaches. For practitioners, the practical lesson is that staying compliant will increasingly mean tracking a living map of obligations across multiple domains—data protection, sexual‑offence law, antitrust and more—where seemingly minor corrections to citations and scope labels can translate into very real changes in who is exposed to liability.