The EU Council moved first.
It agreed an approach to streamline the EU AI Act’s requirements, proposing adjustments to the compliance timeline, a development analyzed by Burges-Salmon and covered in our earlier brief on the Council’s December 2027 and August 2028 proposals. That brief is the starting point for readers new to this story.
Then the Parliament moved.
On March 16, 2026, the European Parliament’s Internal Market Committee (IMCO) and Civil Liberties Committee (LIBE) published their own proposals. The Parliament’s press release confirms MEP committee support for postponing certain AI Act obligations. According to the MEP committee proposals, the new deadlines would be December 2, 2027, for listed high-risk AI systems and August 2, 2028, for high-risk AI embedded in products covered by EU sectoral safety legislation. These mirror the EU Council’s direction closely.
Two institutions. Same destination.
What the EU Council Proposed vs. What Parliament Committees Added
The Council’s approach focused on streamlining, reducing compliance burden, rationalizing the timeline. The Parliament committees largely align on the deadline extensions but add something new: a proposed prohibition on “nudifier” AI systems, which generate non-consensual explicit images of real people.
This matters for a simple reason. The Council’s proposals were about when. The MEPs’ additions are partly about what. A new category of prohibited AI practice changes the scope of compliance work, not just its schedule. Organizations building or deploying image-generation AI need to assess whether their systems fall within this proposed prohibition category, which would carry the regulation’s highest penalty tier.
Under the EU AI Act, penalties for prohibited AI practices reach €35 million or 7% of global annual turnover. That’s a different penalty bracket from the €15 million or 3% that applies to high-risk AI violations, per Baker Botts’ analysis. Don’t conflate them. The nudifier ban, if enacted, would sit in the prohibited category.
The Watermarking Disagreement
Not everything is aligned. On AI-generated content watermarking, MEPs and the European Commission are proposing different deadlines. MEPs have put forward November 2, 2026. The Commission’s proposal is February 2, 2027. That three-month gap isn’t trivial for compliance planning, the two dates fall on opposite sides of most annual planning cycles.
This disagreement has to close before any amendment is final. The June vote will reveal whether MEPs or the Commission’s position prevails, or whether a compromise date emerges. For watermarking obligations specifically, planning teams should track this closely.
What “Aligned” Does and Doesn’t Mean
Calling the Council and Parliament “aligned” is accurate but requires precision. Formal trilogue, the EU’s interinstitutional negotiation process, has not concluded. Both institutions have moved in the same direction, which makes final agreement on delays significantly more probable. It doesn’t make those delays certain.
The formal sequence from here, according to Burges-Salmon’s analysis: a full Parliament vote expected in June 2026, with published amendments in July 2026. Until that vote, nothing about the August 2026 deadline is legally changed.
Four Stakeholder Groups and Their Current Position
*Compliance teams at organizations subject to listed high-risk AI obligations* are in the most uncomfortable position. August 2026 is the operative deadline right now. Recalibrating toward December 2027 based on proposals that haven’t passed creates real compliance exposure if the vote is delayed, narrowed, or fails. The conservative read is to continue August 2026 preparations while designing them to be pausable.
*Providers of AI systems in EU-regulated sectors*, medical devices, machinery, automotive, face the August 2028 proposed extension for sectoral high-risk AI. The longer runway is more practically significant here because embedded AI compliance often requires hardware redesign cycles that can’t be compressed. If the extension passes, that extra time has genuine value.
*AI image and video generation companies* face the nudifier ban proposal regardless of deadline discussions. This is a new prohibition category, not a scheduling matter. Organizations in this space should be assessing their product portfolios against the proposed prohibition’s likely scope now, not after the June vote.
*Organizations with AI-generated content watermarking obligations* need to treat the November 2026 and February 2027 proposals as a range, not a date, until the June vote resolves the MEP-Commission disagreement.
The Decision Compliance Teams Actually Face
The honest framing is this: the probability of delay is now high enough that it changes planning calculus, but not high enough to justify stopping August 2026 work.
The practical path is to continue compliance programs with an eye toward which elements are genuinely deadline-sensitive and which can be paused without consequence if the June vote confirms the delay. Document risk assessments now. Establish a June 2026 decision checkpoint. Don’t conflate “likely delayed” with “definitely delayed.”
The June vote is the next firm date. That’s where the uncertainty resolves.