What the AI Office expects beyond the text of the Act
The director of the European AI Office has described a supervisory practice that asks more of providers than the enforceable text alone requires.
Blog
Blog
Analysis and commentary on legal engineering, automation and the regulation of emerging technology.
Copying Westlaw headnotes to train an AI was not fair use
The Third Circuit has affirmed that ROSS Intelligence infringed copyright by using 2,243 Westlaw headnotes to train its AI legal search engine, and that the use…
The AI Board turns to the machinery of enforcement
The AI Board met on 17 September 2026 and placed market surveillance, conformity assessment and the transparency duties in force since 2 August at the centre…
When a GDPR breach earns a reprimand instead of a fine
The EDPB adopted Guidelines 04/2026 on 17 September 2026. They set five steps for deciding whether a GDPR breach draws a fine, a reprimand or another…
Two new AI Act prohibitions apply from 2 December 2026
Regulation (EU) 2026/1744 inserted two prohibitions into Article 5 of the AI Act. They bind from 2 December 2026 and carry fines of up to 7…
Deleting the data did not discharge the duty to reply
The CNIL fined EXTIA 300,000 euros on 21 July 2026 after 12 erasure requests went unprocessed and 166 people were never told what had happened to…
The first requests for information under the AI Act
The AI Office has sent its first requests for information to providers of general-purpose AI models, and a July implementing regulation sets out what the Commission…
A 24-hour clock now runs for makers of digital products
From 11 September 2026 the EU Cyber Resilience Act obliges manufacturers to report actively exploited vulnerabilities within 24 hours, through a single ENISA platform.
A California bill draws a line around the practice of law
SB 574 would add section 6068.1 to California's Business and Professions Code, forbidding an attorney to delegate the practice of law to generative AI and requiring…
Watermarks in practice, agents by definition
Anthropic sets out how Claude’s text will be marked, the drafters answer the claim that watermarking degrades writing, and the Commission explains where AI agents already…