One year ago, on 19 November 2025, the European Commission proposed the Digital Omnibus: a package meant to „simplify“ EU digital rules. That word was used to justify some of the biggest changes to online data protection since the GDPR went live in 2018. They included proposals such as:
- A narrower definition of personal data
- Looser cookie consent rules
- A new rule that would let companies train and run AI on your personal data without asking you
And this isn’t only about data collected from now on: critics warn it could also open up the posts, chats and documents companies have gathered about you over the years. That resembles a lot what US Big Tech companies have been asking for, as noted by MEP Michael McNamara at a European Parliament hearing
One year later, part of the package is law. But the most important fight, over your data and AI, is still open.
The Digital Omnibus: Two halves, two speeds
The Omnibus came in two parts. The AI part delayed and loosened rules in the EU AI Act. It moved fast and entered into force on 27 July 2026. The data part rewrites the GDPR, and it is still being negotiated. That is where the question of AI training on your data will be decided.
They promised you can always opt out…
Under the GDPR, companies need a legal basis to use your personal data. One of them is „legitimate interest“. This means a company may use your data without consent if its interests outweigh yours. The Commission proposed a new article confirming that companies can rely on legitimate interest to develop and run AI. In return, it promised you an unconditional right to opt out.
… but then they changed their mind: overriding legitimate interest
In October, EU governments agreed on their own version. It keeps the AI article, adding that AI development is not automatically a legitimate interest. But it removes the unconditional right to object. You would keep only the ordinary right, which a company can override if it shows compelling legitimate grounds.
The privacy group noyb warns that, under the government’s text, processing personal data „in the context of AI“ would automatically be lawful, letting Big Tech use personal data collected over decades with almost no restrictions.
The EU’s own data protection authorities, the European Data Protection Board (EDPB) and European Data Protection Supervisor (EDPS), also criticized the AI article. In a joint opinion, they called the AI article unnecessary as it doesn’t really clarify anything.
AI training under the GDPR: Business at the table
Corporate Europe Observatory and LobbyControl counted 655 lobbying meetings with Members of the European Parliament on the Omnibus since November 2025. 83% of them were with business interests. In Denmark, Sweden and Ireland, former government officials have gone on to lobby for Meta on the file.
Bram Vranken, campaigner and researcher at Corporate Europe Observatory, added: „Our research shows that Big Tech enjoys unprecedented access to the highest levels of government across Europe.“
Some wins for digital rights
Pressure from civil society, regulators and MEPs has already changed the package:
- EU governments dropped the narrower definition of personal data which the data protection authorities warned would significantly shrink the GDPR’s reach.
- They restored the GDPR’s protections against fully automated decisions (Article 22).
- In the AI part, the plan to let companies skip registering AI systems they judge „not high-risk“ was stopped, though the registration was watered down.
But the AI training article survived, and so the right you have as a consumer to say no was weakened.
What happens next for the Digital Omnibus
The European Parliament hasn’t made a final decisition on its position. The two lead committees for industry (ITRE) and civil liberties (LIBE) have received more than 1,000 amendments and haven’t set a vote date. A vote might happen in early 2027, and only then can Parliament and EU governments negotiate the final text.
That means there is still time to be heard. If you live in the EU, contact your MEP and tell them you want an unconditional right to object to AI training on your data. The lead negotiators are Aura Salla (EPP, Finland) and Marina Kaljurand (S&D, Estonia), but you can also contact your own MEP.
Keep your data out of someone else’s AI
Whatever the final text says, it decides what companies can do with data they already hold about you. The surest way to keep your documents, chats, and email out of someone else’s training set is not to hand them over in the first place.
With Nextcloud, organizations run AI on their own servers. Nextcloud Assistant summarizes mail threads, helps with writing, translates, and answers questions about your documents using local models. Nothing is sent to external AI services unless an administrator turns that on.
If you do want to connect an external provider, the Nextcloud Ethical AI Rating shows for every AI feature how open its software, model, and training data are, so you can decide what fits your data. Self-hosting alone doesn’t make anyone GDPR-compliant, but it does mean you decide who touches your data.