← Back to Insight
AI Governance · EU AI Act

The Organisations Most at Risk Under the EU AI Act Are Not the Ones That Moved Too Slowly

On 2 August 2026, high-risk AI system obligations become enforceable across the EU. The organisations most exposed are those that deployed AI at pace and assumed governance could follow.

A
Abdulwahab B. Adesanya Barrister-at-Law · LLM IP & IT Law (UCD)
June 2026  ·  8 min read

On 2 August 2026, the obligations under the EU AI Act applicable to high-risk AI systems become fully enforceable. Market supervisory authorities across the European Union will, from that date, have the power to require compliance documentation, impose corrective measures, and levy administrative fines of up to €30 million or 6% of global annual turnover — whichever is higher.

The organisations most exposed to that liability are not those that chose caution. They are those that chose speed without governance.


The Evidence Is Unambiguous

IDC's July 2025 research, drawing on responses from over 1,000 enterprise IT decision-makers globally, identified the following:

65%
of enterprises say AI workload growth is their primary driver of infrastructure investment over the next two years
85%
characterise generative AI as a transformative corporate workload demanding commensurate, sustained investment

These figures confirm that enterprise AI adoption is not a future trend. It is a present operational reality. What they also confirm, read against the regulatory backdrop, is that a large proportion of that deployment is proceeding without the governance architecture the law now requires.

IDC's research identifies compliance gaps, inadequate security controls, and the absence of data sovereignty frameworks as material causes of AI initiative failure at scale. These are, without exception, governance failures — not engineering ones. And unlike engineering deficiencies, they cannot be remediated after the fact without regulatory consequence.


What the Law Actually Requires

The EU AI Act does not prohibit ambition. What it prohibits is the deployment of high-risk AI systems without demonstrable, documented, and operational governance.

High-risk systems — defined under Annex III to include AI used in employment and worker management, access to essential services, biometric identification, law enforcement, and the administration of justice — must, from 2 August 2026, satisfy obligations across six core domains:

Core obligations for high-risk AI systems

Art. 9
Risk Management System

Established, implemented, documented, and maintained throughout the AI system's lifecycle. A policy document drafted for the occasion will not suffice.

Art. 10
Data Governance

Management practices addressing training, validation, and testing data — their relevance, representativeness, and freedom from known errors and biases.

Art. 11
Technical Documentation

Prepared before the system is placed on the market or put into service, demonstrating compliance to the satisfaction of national authorities.

Art. 13–14
Transparency & Human Oversight

Natural persons must be in a position to understand, monitor, and where necessary override or halt AI system outputs.

Art. 15
Accuracy, Robustness & Cybersecurity

High-risk systems must achieve appropriate levels of accuracy, robustness, and cybersecurity throughout their operational lifecycle.

Each of these obligations requires institutional infrastructure: documented processes, defined accountabilities, trained personnel, and governance bodies with authority to act. None of them can be satisfied by a legal opinion, however well-drafted.


Governance Is Not a Compliance Function. It Is a Structural One.

The organisations navigating this well are not treating August 2026 as a compliance project assigned to their legal team. They are treating it as the deadline by which their AI governance architecture must be operational — a central function that aligns legal, data, IT, and business teams around documented policies and shared accountability.

"The organisations that treat the Act as a technical specification to be met at the last moment will find, when supervisory authorities begin their reviews, that the documentation they produce does not reflect the reality of how their systems operate."

That gap — between stated governance and actual practice — is precisely what regulators are trained to find.


Three Steps That Cannot Be Deferred

For any organisation operating or procuring AI systems in the European Union, three actions require immediate attention:


August 2026 Is Not the End of This Process. It Is the Beginning of Enforcement.

The EU AI Act does not sunset. The obligations that apply from August 2026 are ongoing. The organisations that invest now in governance architecture that is proportionate, documented, and genuinely operational will be better placed not only to satisfy immediate regulatory demands but to deploy AI at scale — with the confidence that comes from structural compliance, not hope.

Understand your exposure before August

Adesanya AI Advisory offers a focused AI governance health check — a structured assessment of your AI use, risk classification obligations, and the governance gaps that require priority action.

Book a free health check Or contact directly: abdulwahab@adesanyaaiadvisory.com

Source: IDC InfoBrief, "Generative and Agentic AI-Ready Infrastructure Strategies," sponsored by Amazon Web Services and NVIDIA, July 2025 (IDC #US52614724). Statistical data cited as published; analysis and legal interpretation are those of Adesanya AI Advisory.