Enforcement begins
From 2 August the AI Office and national authorities are enforcing, transparency duties apply, and there are three channels through which a breach can be reported.…
Blog
From 2 August the AI Office and national authorities are enforcing, transparency duties apply, and there are three channels through which a breach can be reported.…
Regulation (EU) 2024/1689 classifies parts of justice-sector technology as high-risk and applies to providers outside the Union. For third-country legal service providers this is not a…
The Commission declared the transparency Code adequate. In the same weeks, OpenAI reported an AI agent escaping a laboratory test by hacking external infrastructure.
Rule-based systems give certainty and cannot improvise. Language models improvise and cannot guarantee. For legal documents the interesting answer is usually neither one alone.
423 votes to 57 fixed the new high-risk timetable. The more interesting development was a tool letting people declare, in a format machines can read, what…
Four articles turn the AI Office from a coordinating body into a regulator that can demand documentation, commission its own evaluations, pull a model from the…
Europe is building compute at scale while the unit supervising the riskiest models plans for about forty people. The gap between those two numbers is the…
Council and Parliament have each adopted a position on simplifying the AI Act. High-risk obligations slip to December 2027, while the list of banned practices gets…
A lecture about hashing leaves a law student able to repeat a definition and unable to say what would go wrong. Two small open-source modules changed…
The Commission has put a date on an obligation that reaches ordinary document tools, not only image generators. Feedback on the second draft closes 30 March…