Compliance teams tracking EU AI Act timelines should hold their current implementation schedules. The European Parliament has voted on a package of amendments, part of the Digital Package on Simplification, that would extend key high-risk AI compliance deadlines. Those amendments have not yet been formally enacted as amended official text. Parliament’s vote is a significant procedural step. It is not the same as a changed law.
The distinction matters enormously for organizations mid-implementation. Under the EU AI Act as currently enacted, the original applicability dates remain in force. Reports indicate Parliament’s amendments would move the compliance deadline for general high-risk AI systems to approximately December 2027, and to approximately August 2028 for AI embedded in products already covered by EU product safety laws, medical devices, toys, and similar regulated categories. A November 2026 deadline for watermarking AI-generated content has been reported as part of the same legislative package. These figures come from secondary reporting; the underlying source documents require confirmation before they should be used as planning benchmarks.
A separate vote, on a ban covering non-consensual sexual deepfakes, passed by an overwhelming majority. This is the enforcement signal in an otherwise deadline-extension story. Parliament chose to move forward aggressively on personal dignity protections at the same moment it extended technical compliance timelines. That combination tells a story about where EU political consensus actually lies: the ambition is intact; the implementation pace is being calibrated to practical reality.
The extension proposal stems from a straightforward problem. The technical standards that high-risk AI operators need to demonstrate compliance simply weren’t ready by the original target dates. Extending deadlines is not a retreat from the regulation’s scope, it’s an acknowledgment that the compliance infrastructure must exist before enforcement is meaningful. Organizations that have been building toward the original dates are ahead, not penalized. Their work remains valid; the finish line has moved, not the direction.
The watermarking requirement, applying to AI-generated audio, images, video, and text, carries a reported November 2026 compliance date. If that date is confirmed, it arrives well ahead of the high-risk system deadlines and deserves separate attention. Content teams and marketing organizations deploying generative AI tools may face this requirement before their AI governance colleagues have worked through the high-risk classification framework.
What to watch
the formal legislative process following Parliament’s vote. EU amendments must complete additional procedural steps before they take effect as amended official law. The EU AI Act’s original text, including its original applicability dates, remains the operative compliance document until that process concludes. Organizations should monitor EUR-Lex and the European Parliament’s official channels for formal enactment updates.
The TJS EU AI Act News March 2026 Digest covers these developments alongside the broader amendment timeline. The core compliance planning discipline remains the same: track the official text, not the reporting cycle. When the amendment is formally enacted, update your timelines. Until then, hold your current schedule and build in contingency for the proposed new dates, because Parliament’s vote signals those dates are the likely direction.