The EU AI Act's August 2, 2026 Milestone: What It Means for CE Marking AI-Enabled Instruments
August 2, 2026 was the date most compliance teams had circled for the EU AI Act. It’s a milestone: the bulk of the Act now applies, national authorities and the European Commission's AI Office have begun enforcing it, and the transparency obligations are in force.
But six days earlier, on July 27, 2026, the Digital Omnibus on AI (Regulation (EU) 2026/1744) entered into force and moved the deadlines that matter most to manufacturers of CE-marked instruments. The demanding "high-risk" obligations that reshape conformity assessment were not switched on in 2026. They were pushed to December 2027 and August 2028.
For engineers and product managers building AI features into measurement, control, laboratory, and IT or multimedia equipment, that distinction matters more than either deadline alone. The pressure eased, but the work didn’t go away, and a good deal of it depends on the EMC and product safety foundation you’re already responsible for.
This article breaks down:
-
Which AI Act obligations became enforceable on August 2, 2026, and which ones the Digital Omnibus pushed to December 2027 and August 2028.
-
A two-part test for whether your AI-enabled instrument counts as a high-risk AI system at all, and why your product's CE legislation decides the answer.
-
How the AI Act folds into your existing CE conformity assessment under a single CE mark, and how to use the runway between now and 2028.
The AI Act Sits on Top of CE Marking, It Doesn’t Replace It
The AI Act, Regulation (EU) 2024/1689, entered into force on August 1, 2024 as the EU's first horizontal law for artificial intelligence. It sorts AI by risk and layers obligations onto systems already regulated elsewhere, rather than creating a separate marking regime.
Four sets of rules can apply to the same product independently:
-
Prohibited practices
-
High-risk requirements
-
Transparency duties
-
Rules for general-purpose AI models
An AI-enabled instrument can fall under one, several, or none of these, depending on what the AI does and how the product reaches the market.
That layered design is why the AI Act connects so directly to your existing work. The way AI features interact with EMC and safety testing determines both how the product behaves under test and where it lands in the Act's risk structure.
What August 2, 2026 Triggered, and What It Didn't
Three things became operative around the August 2, 2026 date.
-
The general application of the Act began.
-
Enforcement and penalties came online.
-
Article 50 transparency obligations started to apply.
The penalty ceilings under the Act can reach fines of up to 35 million euros or 7 percent of worldwide annual turnover, and high-risk non-compliance up to 15 million euros or 3 percent, whichever is higher.
What didn’t arrive on that date is the high-risk compliance machinery: risk management systems, data governance, technical documentation, human oversight, and the AI-specific conformity assessment. The Digital Omnibus deferred those obligations, so for the products most GME clients build, they are a 2027 and 2028 concern, not a 2026 one.
The Revised Compliance Timeline at a Glance
These revised dates come from the Digital Omnibus. You can cross-check the framework against the European Commission's implementation timeline, and confirm the deferred dates against the Omnibus text itself before you rely on them for planning.
Is Your AI-Enabled Instrument a "High-Risk AI System"? The Two-Part Test
Most instrument makers assume that adding AI automatically makes a product high-risk. It does not. Under Article 6(1), an AI system is high-risk through the product route only when two conditions are both true.
First, the AI must be a safety component of a product covered by the Union harmonization legislation listed in Annex I, or must itself be such a product. Second, that product must be required to undergo third-party conformity assessment under that legislation.
AI intended solely for non-safety functions no longer counts as a safety component, including AI used only for:
-
User assistance
-
Performance optimization
-
Service efficiency
-
Automation
-
Convenience
-
Quality control
Many AI features in measurement and control equipment fall squarely into that carve-out, which keeps them outside the high-risk category even when the product is otherwise in scope.
Why Your CE Legislation Decides Whether You're in Scope
Whether your instrument can even reach the high-risk product route depends on which CE legislation applies to it, and that’s a question of product category, not marketing language.
Most bench instruments carry the CE mark as measurement, control, and laboratory equipment under the EMC Directive (2014/30/EU) and the Low Voltage Directive (2014/35/EU), typically against EN 61326-1 and EN 61010-1. Neither of those directives is listed in Annex I of the AI Act, and both allow manufacturer self-declaration rather than mandatory third-party assessment. An instrument marked only under EMC and LVD therefore does not reach the Annex I high-risk route at all.
The picture changes the moment you add a radio. A wireless module brings the product under the Radio Equipment Directive (2014/53/EU), which absorbs the EMC and safety objectives and is listed in Annex I. Machinery and medical devices are in Annex I as well. Being covered by an Annex I law is necessary but not sufficient: the product still has to be one that requires third-party conformity assessment, and the AI still has to be a safety component, before the high-risk obligations attach.
The RED already stacks obligations onto wireless products, including the cybersecurity essential requirements enforced through EN 18031. The AI Act adds another potential layer on top of that, which is why understanding your exact regulatory footprint before you design in AI is worth the effort.
One Product, One CE Mark: How the AI Act Integrates With Your Assessment
When an AI system is both high-risk and embedded in an Annex I product, the Act doesn’t create a second mark or a parallel file. Article 48 provides for a single CE marking that signifies conformity with both the AI Act and the underlying product legislation.
Article 43 handles the mechanics. For products already covered by Annex I sectoral law, the manufacturer follows that law's existing conformity assessment procedure, and the AI-specific requirements become part of that same assessment. Where a notified body is already involved under the sectoral legislation, that body can control conformity with the AI requirements too.
The AI conformity assessment builds on the product conformity work you already do, so the accuracy of your EMC results, your safety evaluation, and your RED essential-requirements evidence feeds directly into the AI layer.
What the Transparency Obligations Mean for Your Products Now
Article 50 is the piece that’s live today, here’s what it means. It requires:
-
That people be told when they’re interacting with an AI system.
-
That AI-generated or manipulated content be marked in a machine-readable way.
-
That emotion-recognition and biometric-categorization systems disclose their operation to affected people.
For most bench measurement and control instruments, Article 50 has a narrow footprint, since these products rarely converse with users or generate content. The exposure is higher for IT and multimedia equipment that ships with AI chat interfaces, voice assistants, or generative features, where a disclosure or labeling obligation can apply from day one.
The safe approach is to map your product's AI behavior against Article 50 now, rather than assume it’s exempt. The obligation is modest to meet when designed in early, and awkward to retrofit after launch.
How to Use the Runway Between Now and 2028
The deferral is time to build, not time to stand down. The obligations that were pushed back are the ones that take the longest to put in place, and they still rest on a conformity foundation that has to be correct first.
A practical sequence for the runway looks like this:
-
Confirm which CE legislation applies to your product and whether any of it sits in Annex I.
-
Classify each AI function against the safety-component test and the Annex III use-case list.
-
Check Article 50 transparency exposure.
-
Keep your EMC, safety, and RED evidence current and defensible.
Planning a single, well-scoped EMC campaign that supports CE, UKCA, and FCC at once keeps that foundation efficient as the AI requirements phase in.
Manufacturers who treat the extra time as engineering runway, rather than a reprieve, will reach 2028 with an integrated file instead of a scramble.
Turning the AI Act Timeline Into a Strategic Advantage
The August 2, 2026 milestone didn’t land on CE-marked instruments the way early headlines suggested, but it did make one thing clear. The AI Act now sits on top of the EMC, safety, and radio compliance work that already determines whether your product reaches the European market.
Getting that foundation right now is what makes the AI layer manageable when it phases in through 2028. GME's accredited EMC and product safety testing gives you the accurate, defensible results the integrated conformity assessment is built on, and the front-end guidance to plan for it before you design in AI.
Talk to a GME Engineer about where your AI-enabled instrument fall in the AI Act's risk structure and how to keep its CE foundation solid.