Every EU and Irish law, directive and standard that touches AI, on one page. What already applies, what is coming, who it is for, and what it means in plain English.
Who: Anyone whose AI tools touch personal data
If an AI tool sees personal data, GDPR applies. People can object to important decisions made only by a machine (Article 22).
Who: Any organisation
The information security management standard. The base the AI security guidance (ISO/IEC 27090) builds on.
Who: Anyone wanting a free, practical AI risk method
A free US framework (govern, map, measure, manage) that many EU firms use alongside the AI Act.
Who: Risk and compliance teams
How to run risk management for AI, built on ISO 31000.
Who: Organisations that use or build AI
A management system for using AI responsibly. The one AI standard you can be certified against.
Who: Online platforms and search engines
Platforms must explain how their recommendation algorithms work, and the largest must assess their risks.
Who: Anyone who builds, sells or uses AI in the EU
The core AI law. It sorts AI by risk: banned, high-risk, transparency duties, or minimal. The duties start on the dates below.
Who: Medium and large firms in 18 sectors, and their suppliers
Cyber security duties, with management boards personally accountable. The EU deadline has passed; Ireland's National Cyber Security Bill is still being drafted.
Who: Banks, insurers, investment firms and their tech suppliers
Financial firms must manage the risk from their technology suppliers, AI providers included.
Who: Every organisation using AI
Some uses are banned, such as social scoring and emotion recognition at work. Organisations must support AI literacy for staff who use AI (Article 4).
Who: Was aimed at people harmed by AI
The Commission withdrew it. Claims for harm from AI now run mainly through the new Product Liability Directive (Dec 2026).
Who: Teams assessing AI systems; certification bodies
42005: how to assess an AI system's impact on people. 42006: the rules for bodies that certify ISO 42001.
Who: Makers of large AI models
Model makers must document their models, respect copyright and publish a training-data summary. A voluntary Code of Practice (Jul 2025) shows how.
Who: Anyone using cloud services or connected products
Rights to the data your connected products create, and the freedom to switch cloud provider.
Who: Governments, and in time public bodies
The first binding international AI treaty, on human rights and democracy. The EU ratified it in 2026; it needs more ratifications to take effect.
Who: Providers of high-risk AI
The first European standard written to show AI Act compliance, for quality management. More are being drafted.
Who: Anyone using or selling AI in Ireland
How Ireland enforces the AI Act: existing regulators police AI in their own sector (for example the Central Bank, the HPRA and Coimisiún na Meán), coordinated by the new AI Office of Ireland.
Who: Anyone running chatbots or making AI content
Tell people when they're talking to an AI. Label deepfakes. Mark AI-generated content so machines can detect it (Article 50).
Who: Makers of software and connected products
Report actively exploited vulnerabilities and severe incidents within 24 hours, for products already on the market too.
Who: Anyone building or running AI systems
Security threats specific to AI, such as data poisoning, prompt injection and model theft, and how to defend against them.
Who: Makers of AI that generates images, video or audio
AI that generates child sexual abuse material or non-consensual intimate images is banned. Generative systems sold before Aug 2026 must mark their output too.
Who: Anyone who makes or sells products, software included
Software and AI systems count as products, so people harmed by a defect can claim compensation. Applies to products placed on the market from this date.
Who: Anyone using cloud services
Cloud providers can no longer charge you to move your data to another provider.
Who: Providers and users of AI in hiring, credit, education, essential services and more
Risk management, human oversight, documentation and registration for high-risk AI (Annex III). Delayed from 2026, not cancelled.
Who: Makers, importers and sellers of digital products
Security by design, security updates for the support period, a software bill of materials (SBOM), and CE marking.
Who: Makers of products such as medical devices and toys that contain AI
High-risk rules for AI that acts as a safety component in products already regulated by EU law (Annex I).
Guidance only, not legal advice. Dates are when a duty starts to apply, not when a law was passed. Standards are voluntary; only ISO/IEC 42001 and ISO/IEC 27001 can be certified. Spotted something out of date? Tell me.
Sources
Most organisations need three or four of these, not all of them. Take the free readiness check, or book a call and we'll work out your list together.