If your understanding of the EU AI Act was that the tough high-risk rules for finance hit in August 2026, that understanding is now out of date. The 2026 Digital Omnibus package, which moved through the European Parliament and Council in June, deferred the high-risk obligations most relevant to finance, pushing that regime out to late 2027.

That is a meaningful change, and it has been widely misreported as either a cancellation or a non event.

Neither framing is right. The high-risk duties moved later, but transparency duties under Article 50 still apply from August 2, 2026, with a short grace period for systems already on the market.

So a finance team needs a clear picture of the new two speed timeline: what it must do now on transparency, and what it should prepare for on high-risk. This piece lays that out and explains why the deferral is best used as preparation time.

EU AI Act
Transparency rules
High-risk systems
Compliance timeline
Governance
Audit trail
Regulatory prep

What actually changed

The original AI Act timeline had high-risk obligations phasing in on a fixed schedule, with the categories most relevant to finance, such as creditworthiness assessment and risk pricing in insurance, among them. The 2026 Digital Omnibus package restructured that schedule.

The high-risk obligations were deferred, and the regime for those finance relevant categories is now expected to apply in December 2027, with certain other high-risk categories moving even later.

What was not deferred is the transparency layer. Article 50 obligations, which require disclosing when a person is interacting with an AI system and labeling AI generated or manipulated content, still take effect on August 2, 2026.

Systems already on the market when the rules apply are given until December 2, 2026 to comply. So the practical result is a split timeline rather than a single moved date, and confusing the two parts is the main risk right now.

A staged timeline view, illustrating transparency duties now and high-risk duties later

What applies to finance teams now

The obligations that bite from August 2026 are the transparency ones, and they are lighter than the high-risk regime but not trivial. If your finance function deploys AI that interacts with people or generates content, you need to be able to disclose that AI is involved and to label AI generated or manipulated material where the rules require it.

That is a documentation and disclosure discipline more than an engineering overhaul, but it still needs to be in place, and the grace period for existing systems runs only to December 2026.

The practical step is to inventory where AI touches anything customer facing or content producing in your function, and to make sure the required disclosures and labels are handled.

This is exactly the kind of task that is easy to defer and awkward to retrofit under time pressure, so doing it now, while it is the only live obligation, is the low stress path.

Inventory where AI in your function interacts with people or generates content.
Ensure required disclosure that AI is involved, and labeling of AI generated or manipulated content.
Handle existing systems within the grace period that runs to December 2, 2026.

What to prepare for on high-risk

The deferral of the high-risk regime to late 2027 is real breathing room, but it is breathing room to prepare, not to ignore. When the high-risk obligations do apply to finance relevant uses, they are substantial: risk management systems, data governance, documentation, human oversight, accuracy and robustness, and logging.

Those are not things you assemble in the final quarter before a deadline. They are properties you build into how AI is deployed from the start.

The teams that will meet the 2027 regime comfortably are the ones treating the extra time as a runway. Building AI deployments that are governed, documented and auditable now means the high-risk obligations, when they arrive, are largely already satisfied by how you work rather than a scramble bolted on at the end.

The deferral rewards preparation and punishes procrastination, exactly as such deferrals usually do.

The deferral is a runway, not a reprieve. Build governed, auditable AI now and the 2027 high-risk regime is mostly a matter of documenting what you already do.

Why governed AI satisfies both timelines

There is a reassuring unity beneath the two speed timeline. The same foundation that handles the transparency obligations today also positions you for the high-risk regime later: knowing where AI is used, keeping humans in control of consequential decisions, documenting what the AI does, and maintaining a record that can be examined.

Transparency asks you to disclose and label. High-risk asks you to govern, oversee and log.

Governed AI does all of it as a matter of course.

This is why the right response to the split timeline is not to build two separate compliance projects on two different schedules. It is to build governed AI once, which satisfies the transparency duties now and grows naturally into the high-risk requirements as they arrive.

A team that governs its AI properly is not tracking two deadlines with dread. It is doing the thing that both deadlines ultimately ask for.

A note on scope and jurisdiction

One reason the AI Act matters even to teams that do not think of themselves as European is its reach.

The obligations can apply based on where an AI system's output is used, not only on where the provider or deployer is based, which means a finance function serving European customers or operations can fall within scope even if its headquarters sit elsewhere.

That is why treating the Act as somebody else's problem is risky, and why understanding which of your uses might be in scope is worth doing before either deadline forces the question.

The practical response is to map your AI uses against the Act's categories at a high level: which are purely internal and low risk, which involve the transparency triggers that apply now, and which would fall into the finance relevant high-risk categories when that regime lands in 2027.

You do not need a definitive legal classification to start, and this is exactly the kind of question to confirm with qualified advisors. But a rough map tells you where to focus, and it turns a vague sense of regulatory exposure into a short, concrete list of things to watch and prepare.

Uncertainty about scope is itself a reason to build governed, documented AI, because governed AI is defensible whichever way a scope question resolves.

The Act can apply based on where AI output is used, so non European teams can still be in scope.
Map your uses at a high level: internal and low risk, transparency triggering now, or high-risk later.
Confirm classification with advisors, and build governed AI so you are defensible whichever way scope resolves.

Where ChatFin fits

ChatFin applies AI to finance work with human approval on every write and a full audit trail on every action, which is the governed foundation both parts of the AI Act timeline reward.

The audit trail and human oversight map directly onto the high-risk regime's oversight and logging expectations, and the disciplined record of where and how AI is used supports the transparency obligations that apply now. You build governed AI once, and it serves both the near term and the 2027 requirements.

So the moved deadline is not a reason to relax and it is not a reason to panic. It is a reason to build AI the governed way, which is what ChatFin does, and which turns a two speed regulatory timeline into a single, steady discipline.

The deadline split, it did not vanish. Build governed AI once and meet both timelines.

ChatFin applies AI to finance work with human approval and a full audit trail, the governed foundation that satisfies the AI Act transparency obligations now and positions you for the high-risk regime in 2027. One governed approach, both timelines covered.

The teams that use the deferral to build governed, auditable AI will meet the future regime without a scramble. The ones that treat it as a reprieve will meet it under pressure.

Handle transparency now, prepare for high-risk on the new timeline, and build governed AI once to cover both.

Book a Demo

This article is general information for finance professionals and is not legal, accounting or tax advice. Rules, standards and frameworks change and apply differently by situation, so confirm how any of them affects your organization with your own qualified legal, accounting and compliance advisors.

Related Articles