The European Commission’s new “digital omnibus” proposal has shaken up the conversation around the EU AI Act. Headlines talk about delays, simplification measures and lighter obligations for companies building or deploying AI. After years of intense legislative negotiations, this latest package has triggered mixed reactions: business associations welcome a more realistic timeline, while digital-rights organisations warn of a regulatory rollback.
But what do these proposed changes actually mean in practice? And more importantly — should companies now slow down their AI governance efforts? The short answer: absolutely not.
The Commission’s adjustments are far from a dismantling of the EU AI Act. They simply acknowledge a reality: Europe’s technical standards, notified bodies, and conformity assessment ecosystem are not yet ready for full enforcement in 2026. High-risk AI systems — including recruitment and worker-management algorithms — would receive a deadline extension until December 2027 (and in some cases August 2028). Certain administrative burdens would be reduced, such as the requirement to register “excluded” Annex III systems. Provider obligations around AI literacy might shift partly to the Member States. And a GDPR-related change would allow limited processing of special-category data for the purpose of bias detection.
These pragmatic adjustments make implementation more realistic. But they do not change the spirit of the law — and more importantly, they do not change the underlying ethical obligations that companies already carry today.
Compliance is not the destination. It’s the baseline.
There is a growing misunderstanding in the market that AI governance is “AI Act compliance”. This is a mistake. Compliance is the minimum. Ethical and fair AI is the goal.
Even if every reference to the EU AI Act disappeared tomorrow, companies building or deploying AI systems — especially in areas that affect people’s livelihoods — would still carry responsibilities rooted in long-standing legal frameworks. Anti-discrimination law, labour law, equal treatment requirements, consumer-protection law, and privacy law have always applied to automated decision-making. The AI Act did not create the concept of fairness, transparency or accountability. It simply operationalised them.
A recruitment algorithm that systematically disadvantages women, older workers or minority candidates is unlawful today — not in 2027. A performance-monitoring model that produces opaque or unexplainable risk scores would violate workers’ rights under existing EU and national legislation. And GDPR’s Article 22, data-minimisation principles, purpose limitation requirements and transparency obligations already limit what companies can do with AI-generated insights.
The core idea is simple: an AI system that is legally compliant is not necessarily ethical — but an ethical AI system will almost always be compliant.
The spirit of the EU AI Act is unchanged
Despite the noise, the underlying vision of the EU AI Act has remained intact. Its purpose is not procedural box-ticking but the protection of fundamental rights. In this sense, the Act aligns with global frameworks such as:
- the OECD AI Principles, emphasising transparency, robustness, fairness and human-centred design,
- the NIST AI Risk Management Framework, centred on trustworthy AI,
- ISO 42001, the world’s first auditable AI Management System standard,
- and the broader family of ISO AI standards — including ISO 23894 (AI risk management), ISO 24027 (AI bias), ISO 25058 (AI quality), and ISO 42119-2 (AI testing).
Across all these frameworks, the north star remains the same: AI should be transparent, accurate, accountable, robust, inclusive and fair.
Deadline shifts do not dilute these principles. If anything, they give organisations more space to embed them properly instead of rushing.
This is exactly the right moment to implement ISO 42001
The extension to 2027 is not a signal to relax — it is a signal to do the job well. The companies that treat the delay as a pause will fall behind. The companies that use this time to professionalise their AI governance will gain a decisive advantage.
ISO 42001 matters here for one reason: it remains the most complete and practical framework for building responsible AI governance. It translates abstract principles into concrete organisational structures: leadership roles, accountability lines, lifecycle processes, risk assessments, data-governance requirements, incident handling, documentation, continuous monitoring and internal audits. It provides the blueprint for operationalising transparency, fairness and robustness.
And strategically, it does something even more valuable: it allows companies to signal trustworthiness to customers, regulators, investors and employees. In hiring technologies, for example — where decisions directly reshape someone’s career and livelihood — trust is everything. Candidates will increasingly expect transparency. Clients will expect explainability, fairness reporting, documented controls and visible risk-mitigation measures. Employees will expect their employer to use AI responsibly and ethically. The companies that can demonstrate all of this will win.
ISO 42001 enables that signal long before the CE-marking process starts.
Trust is a competitive advantage — not a compliance burden
AI is becoming deeply embedded in workflows: recruitment, promotion, workforce planning, performance analysis, customer scoring, security, operations, and countless other domains. In this environment, the greatest risk for any organisation is not an audit finding or a fine — it is losing the trust of the people it relies on.
Trust from customers that your AI will not harm them. Trust from employees that your algorithms will not judge them unfairly. Trust from partners that your models are reliable, well-governed and explainable.
This is why ethical and fair use of AI should remain the north star for every organisation, regardless of compliance timelines. Regulations will evolve, deadlines will shift, and political winds will change — but the need to design AI systems that are safe, fair, transparent and aligned with human rights will not.
The companies that understand this will invest now. The companies that wait will be forced to catch up later — under pressure, and without the trust advantage that early, thoughtful governance creates.
The EU may adjust timelines, but the direction is unchanged. The world is still moving toward trustworthy AI. This is the moment to lead — not delay.
Use the extension well: build an audit-ready AI management system now.