#TwitchLawyer
What’s next? The EU Commission is expected to present the proposal on 19 November. Then the EU Parliament and Member States get their say. This will decide how far “legitimate interest” can go and how much control users really keep. 10/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:40 PM
For gamers, streamers and creators that means more of your content, voice and behaviour data can be fed into AI systems by default. Opt-outs get weaker, transparency fuzzier – and you learn you’re “in the training mix” when models are already live. 9/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:40 PM
This isn’t just theory. If “legitimate interest” becomes the default for AI, it shapes how (e.g.) game platforms, streaming sites and socials train on gameplay, VODs, clips, chats and DMs – from moderation to recommendations to generative tools. 8/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:40 PM
Interest groups talk about “rollback” and “death by a thousand cuts” of the GDPR. Academics warn about clashes with CJEU case law and the EU Charter. Politically it’s sold as red-tape cutting while quietly rewriting AI and tracking rules. 7/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:39 PM
There *are* improvements: harmonised DPIA templates, clearer lists for when impact assessments are needed, plus promises of more transparency and a stronger right to object. Critics see this mainly as risk of more paperwork, not more power for users. 6/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:39 PM
At the same time, the draft narrows what counts as “personal” and “sensitive” data. Pseudonymous IDs, logs and some inferences may fall outside the stricter rules. For AI pipelines built on ad IDs and profiles, that’s a downgrade. 5/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:39 PM
That nudges the default. Today you only touch sensitive data if it’s strictly necessary and fits a narrow exception. Under the Omnibus logic it’s easier to say: “we couldn’t realistically strip this data out without breaking the AI”. 4/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:39 PM
Big shift for AI: training, testing and running models would be labelled a “legitimate interest”. That sounds like firms don’t need consent for training on EU data, as long as a balancing test looks OK on paper. 3/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:38 PM
Officially, the Digital Omnibus is about “simplifying” EU digital law. In reality, the leak goes for core GDPR concepts: what counts as personal data, sensitive data and how strong users’ rights really are. 2/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:38 PM
The EU is flirting with rolling back parts of the strong privacy rules of the General Data Protection Regulation (GDPR). Leaked drafts of the “Digital Omnibus” would make AI training on personal data easier – and protection weaker. Thread 👇 1/10

#TwitchLawyer #GDPR #AI
November 16, 2025 at 1:38 PM
Key takeaways: be mindful of lawful acquisition, keep provenance logs, set retention limits, respect EU opt-outs, and expect more licensing. Keep an eye on new US rulings and EU guidance under the AI Act. Thanks for coming to my TED Talk. 10/10

#TwitchLawyer #TechLaw #AI
August 31, 2025 at 11:43 AM
So outcomes can diverge. In the US, training on lawfully obtained books may be fair use. In the EU, a rightsholder's opt-out can close the door, even if the source copy is lawful. In both systems, pirate sources are a no-no. 9/10

#TwitchLawyer #TechLaw #Copyright #AI
August 31, 2025 at 11:42 AM
Those EU exceptions don’t bless illegal sources: copying from pirate sites is still unlawful. The EU’s AI Act also expects general-purpose AI providers to have a copyright-compliance policy and to respect rightsholder opt-outs. 8/10

#TwitchLawyer #TechLaw #Copyright #AI
August 31, 2025 at 11:42 AM
Europe works differently. The EU uses Text and Data Mining exceptions. Research bodies may mine works. Commercial users may mine unless rightsholders have issued a machine-readable opt-out. 7/10

#TwitchLawyer #TechLaw #Copyright #AI
August 31, 2025 at 11:42 AM
The guidance: how you get and keep data matters. Keep provenance (where data came from), avoid retaining unlawful copies, and license content where needed. In short: lawful acquisition plus disciplined retention = very good. 6/10

#TwitchLawyer #TechLaw #AI
August 31, 2025 at 11:41 AM
Important to note: this is one district-court ruling. It's persuasive, but not binding nationwide (and certainly not globally). On 26 Aug 2025 the parties settled, so there will be no appeal. Still, the opinion offers useful guidance. 5/10

#TwitchLawyer #TechLaw #Copyright #AI
August 31, 2025 at 11:41 AM
What the judge said: training on lawfully obtained books was fair use, because it's highly transformative and technically necessary to make the AI system work. He treated that separately from an alleged permanent library of pirated books. 4/10

#TwitchLawyer #TechLaw #AI
August 31, 2025 at 11:41 AM
US basics: “fair use” is an exception to copyright. Courts balance four factors: purpose, nature, amount, and market effect. If the use is transformative and justified, with limited market harm, it can be lawful without permission. 3/10

#TwitchLawyer #TechLaw #AI
August 31, 2025 at 11:40 AM
The lawsuit: A group of authors said Anthropic copied their books, including from pirate libraries, to build and train large language models. The core questions: is training on those texts allowed, and does it matter how the copies were obtained? 2/10

#TwitchLawyer #TechLaw #AI
August 31, 2025 at 11:40 AM
I’ve been reading up on AI model training and copyright law. It isn’t my core field, but here’s what I’ve learned from the Authors v. Anthropic case in the US.👇 1/10

#TwitchLawyer #AI #TechLaw #Copyright
August 31, 2025 at 11:40 AM
Why it matters: if courts can be dragged into the dock over inconvenient rulings, rights become negotiable. Today it’s deportations; tomorrow it could be your business or liberty. Independent courts protect people, not politics. 4/4

#TwitchLawyer #RuleOfLaw #SeparationOfPowers
August 27, 2025 at 5:45 PM
Courts must stay neutral referees. If the executive can sue judges for their rulings, the bench becomes a political target. You need that blocked, so people get a fair, independent hearing. 3/4

#TwitchLawyer #RuleOfLaw #SeparationOfPowers
August 27, 2025 at 5:45 PM
Rule of Law 101: powers are separated. Lawmakers make the law; the executive enforces; courts say what the law is and keep the others in check. You don’t sue judges for judging. You appeal. 2/4

#TwitchLawyer #RuleOfLaw #SeparationOfPowers
August 27, 2025 at 5:44 PM
A federal judge dismissed DOJ’s suit against all #Maryland federal judges, brought to attack a standing order pausing #deportations for 48 hours when an immigrant files a habeas petition. A win for the rule of law. Here’s why. 👇🏼 1/4

#TwitchLawyer #RuleOfLaw #SeparationOfPowers
August 27, 2025 at 5:44 PM
Bottom line: central-bank independence is a public good. Fed officials should be hold to account through law and evidence. Otherwise, markets will rapidly price-in uncredibility, the bill landing on households and jobs. 8/8

#TwitchLawyer #Fed #CentralBankIndependence
August 26, 2025 at 9:40 AM