EU bans AI sexual deepfakes and delays high‑risk AI rules

The EU’s new ban on AI-generated sexual deepfakes and CSAM is one of the clearest normative lines it has drawn in the AI space, while the same reform quietly gives industry more time to comply with onerous high‑risk AI rules. Together, these moves show Brussels trying to balance public outrage over the most egregious abuses with a desire not to choke off broader AI deployment. For companies building powerful generative models, it narrows the space of “anything goes” image generation and raises the bar on content filters and traceability.

From an AGI-race lens, the ban itself is unlikely to slow technical progress: firms advancing cutting‑edge models were already under pressure to curb non‑consensual porn. But the delayed timelines for high‑risk provisions are significant. Extra breathing room until late 2026–2027 makes it easier for European banks, hospitals, and public agencies to experiment with AI systems, potentially increasing demand for advanced models and infrastructure. That, in turn, could pull more frontier and open‑weight capabilities into regulated, high-stakes settings where feedback, logging, and auditing are rich — fertile ground for training and evaluating more capable agents.

The deeper impact may be geopolitical. By visibly cracking down on AI-enabled sexual abuse while softening the compliance cliff for business, the EU is trying to craft a model of “values-forward but industry-friendly” AI governance. How credible that balance proves will influence whether other regions borrow from the EU playbook or seek alternative regimes that are looser or more restrictive.