Europe Delayed Its Toughest AI Rules. The Bigger Question Is Who Gets to Shape Them

For more than a year, compliance teams across Europe and beyond treated August 2, 2026, as a hard deadline. That was the date the European Union’s toughest AI rules covering standalone high-risk systems in hiring, credit, healthcare, and law enforcement under Annex III were set to become legally enforceable. However, negotiators reached an agreement to delay those obligations until December 2, 2027 (with embedded product safety AI pushed to August 2, 2028). The delay, part of a package known as the Digital Omnibus, was formally adopted and entered into force in July 2026, amending the EU AI Act regulatory framework.

What did not move is easy to miss amid the confusion. While general-purpose AI provider obligations took effect earlier in August 2025, transparency duties requiring that people be told when they are interacting with an AI system took effect on schedule on August 2, 2026. However, the obligation to label AI-generated content was delayed to December 2, 2026 for systems already on the market before August 2, 2026. The European Commission, through the EU AI Office, also gained direct powers to investigate and fine noncompliant general-purpose AI providers as of August 2, 2026. In other words, the headline deadline came and went with much of its substance intact, while its most consequential provisions quietly moved to a different calendar entirely.

That mismatch, between what the public expected to change on August 2 and what actually did, is a small but telling window into how AI governance really gets built. It is not a single dramatic law arriving fully formed. It is a running negotiation among regulators, companies, and lobbyists over which obligations survive contact with industry pressure and which get pushed down the road, often with limited public visibility into who is in the room making that call.

Who Actually Shapes These Decisions

That question, of who gets a seat at the table when AI’s rules are written, is one that AI governance strategist Azahara Corrales has spent much of the past year pressing on directly. In an interview and feature published by Lorka AI in August 2026, Corrales argues that regulatory debates tend to focus narrowly on what AI systems can do and where technical limits should apply, while spending far less attention on who builds them, who is trained to use them, and who is in a position to influence how they get deployed noting particularly how gender disparities in AI usage and representation create systemic blind spots.

I would make diversity mandatory at every stage, in who builds the technology, who trains it, who reviews it, who uses it, and who decides what data goes in,” Corrales said. “What one person misses, another will catch.” Her point is not primarily about optics. It is a practical claim about how blind spots get caught before they are built into systems used at scale, and about who is currently positioned to catch them.

Image Source: Pixabay

The Numbers Behind the Governance Gap

On that count, the current data is not encouraging. A LinkedIn analysis covering companies across 24 countries, published in mid-August 2026, found that women with AI backgrounds held just 13 percent of C-suite and top executive roles in AI, compared with significantly higher representation in non-AI sectors, Bloomberg reported. That gap sits squarely inside the group of people most likely to influence exactly the kinds of governance, disclosure, and deployment decisions Corrales describes: the executives setting internal AI policy, approving model releases, and deciding how aggressively to push back on delayed regulatory obligations like the ones that just shifted in Brussels.

The mismatch is not simply about headcount. It is about proximity to decisions. Governance structures, whether inside a single company or across an entire regulatory bloc, are shaped disproportionately by whoever already holds the executive titles, board seats, and lobbying access needed to be in the conversation when a deadline like August 2 gets renegotiated. If that group skews heavily toward one demographic, the resulting rules are more likely to reflect that group’s blind spots, whether or not anyone involved intends that outcome.

Delay Is a Governance Choice, Not a Neutral Pause

It is worth being precise about what the Digital Omnibus delay actually represents. It is not a rejection of AI oversight, and transparency and disclosure obligations for chatbots, deepfakes, and general-purpose models remain in force. But pushing the toughest, most operationally demanding obligations, covering AI used in hiring decisions, credit scoring, and law enforcement, back by 16 months (from August 2026 to December 2, 2027) is itself a substantive governance decision, made under pressure from industry groups arguing that compliance timelines were unrealistic.

Decisions like that are made by people, in specific rooms, weighing specific interests against each other. The composition of those rooms, who is empowered to argue that a deadline is unworkable, and who is positioned to argue the opposite, is not a footnote to the outcome. It is a significant part of what determines it.

What Comes Next Is Still Being Decided

None of this means the questions Corrales raises are unanswerable, or that the current governance gap is permanent. The Digital Omnibus delay itself is a reminder that AI regulation is not a fixed destination but an ongoing negotiation, one that will be revisited well before the new December 2, 2027 deadline arrives. Every one of those revisiting points is another moment where the composition of the room matters.

For now, the practical lesson of August 2026 is less about any single rule and more about the mechanism. The most consequential AI decisions rarely land with the drama of a headline deadline. They land in quieter negotiations over which obligations survive, which get delayed, and who gets consulted along the way.

Governance Isn’t Only Written in Brussels

It is worth noting that formal regulation is only one layer of how AI actually gets governed. Every company deploying a high-risk system still has to decide, internally, how much human oversight to build in, how transparent to be with employees and customers, and how quickly to move once a compliance deadline shifts. Those internal choices are made by the same relatively narrow set of executives the LinkedIn data describes, where women hold just 13 percent of C-suite roles meaning the representation gap at the top of AI companies shapes day-to-day governance decisions long before any regulator gets involved, and largely outside public view.

Corrales points to a project in South Africa where local women built an AI tool allowing individuals in danger to log in and access immediate safety resources as a concrete example of what changes when the people closest to a problem have a hand in shaping the technology meant to address it. It is a small example next to the scale of the EU AI Act, but it illustrates the underlying argument: governance is not only what gets written into law. It is also what gets built, by whom, and in response to whose sense of what actually needs solving.

Whether the process that produces both forms of governance legal and technical becomes more representative over time, or simply keeps running on the same narrow set of voices that have shaped it so far, will likely do more to determine what AI becomes than any single date on a regulatory calendar.

 About Lorka AI

Lorka AI is an all-in-one AI platform that provides access to multiple chat models and productivity tools designed to help professionals streamline workflows, research, and content creation.

Media Contact

Company Name: Lorka AI

Contact Person: Claire Parker

PR Manager

media@lorka.ai

Website: https://www.lorka.ai/