EU AI Act High-Risk Rules Take Effect This Week

The European Union's AI Act moved into its most consequential phase on 2 August 2026. From that date, the obligations for high-risk AI systems started to apply across the bloc. Businesses that build or use AI in hiring, credit scoring, education, or critical infrastructure now carry binding legal duties, not voluntary guidelines.

The timeline was fixed when the Act passed in 2024. The ban on unacceptable practices arrived first in February 2025. Rules for general-purpose AI models such as large language models followed in August 2025. The high-risk tier was always the biggest piece, and it landed this week.

What counts as high-risk

The EU's definition covers AI used in recruitment, employee management, creditworthiness assessments, education and vocational training, and safety components of critical infrastructure. Providers must put in place risk management, data governance, human oversight, and technical documentation. They must also register their systems in an EU database before deployment.

Enforcement is where the teeth are. Fines for prohibited practices can reach €35 million or 7% of global annual turnover, whichever is higher. Most high-risk breaches sit in a lower band of €15 million or 3%. Either way, this is not a regime you ignore.

Why Australian businesses should care

The Act reaches beyond Europe. If your AI system places products or services on the EU market, or affects people in the EU, the rules can apply even when your company is based in Australia. That is less exotic than it sounds. Australian recruitment platforms, fintechs, and software firms regularly serve European customers without realising the compliance switch flipped this week.

For most small businesses the practical message is simpler. The EU is the reference point for where AI regulation is heading, and Australia has no binding equivalent yet. The federal voluntary AI Safety Standard and state-level moves point the same direction. Good data governance and human oversight now means less rebuilding later.

A short checklist for this week

First, list every AI tool you use and ask whether it touches hiring, credit, education, or safety-critical decisions. Second, if you sell into Europe, confirm your AI providers have published their own EU compliance documentation. Third, document how a human can review or override the AI's decisions. That last step is cheap, useful, and required in the EU regardless.

The remaining phases keep coming. High-risk systems embedded in products covered by other EU safety laws follow in 2027. Compliance is not a one-off project, and treating it as a habit is the only approach that survives contact with the real world.

This article was written by Triweb AI's editorial team based on analysis of today's leading AI news sources.