That’s not what’s happening.
The European Parliament just adopted its position on amendments to the AI Act (the so-called AI Omnibus), and the reality is more nuanced:
- The rules will largely stay the same
- The timelines are changing and deadlines are pushed back
- Some risks are now explicitly banned like nudifier apps
Here’s what actually matters if you’re building or using AI:
- Deadlines are pushed back
High-risk AI obligations are likely delayed to 2027–2028. That gives breathing room, but not a free pass. Regulators will still expect mature compliance.
- New explicit bans
AI systems that generate non-consensual nude images (“nudifier apps”) are now clearly prohibited. This is a shift toward stricter enforcement on harmful generative AI.
- Transparency rules stay
If you’re using chatbots or generative AI, disclosure obligations remain. You just get more time to implement them.
- Core obligations are unchanged. This is the part many miss:
- Risk classification stays the same
- High-risk requirements stay the same
- Governance, documentation, and oversight are still required
- AI literacy is already in force and stays in force
Article 4 (AI literacy) is not amended and already applies. Many companies are behind on this.
What does this mean in practice?
If you’re waiting because “the law might change,” you’re taking the wrong approach in our opinion.
This update is not a rollback. It’s a recalibration:
- More time
- Slightly clearer rules
- Same fundamental obligations
The companies that use this extra time to build proper AI governance will be in a much stronger position when enforcement kicks in eventually.
If you’re unsure how this impacts your AI systems, now is the right moment to reassess your compliance roadmap. And if you need AI literacy Training, we’re happy to help with that too.