There's no single EU "AI liability law" — the one that was proposed has been withdrawn. Instead, three frameworks overlap: the revised Product Liability Directive, national tort law, and the AI Act's own obligations, which can become evidence in a claim. Here's how they fit together, and where Article 50 comes in.
If you've read about the EU AI Liability Directive, it would have governed exactly this question — easing the burden of proof for people harmed by AI systems. It won't. The European Commission announced its withdrawal in its February 2025 work programme, citing no foreseeable agreement on the text, and the proposal was formally withdrawn in October 2025.
That leaves the question to the frameworks below.
Directive (EU) 2024/2853 replaces the EU's 1985 product liability rules. Its biggest change for AI is that software — including AI systems — now counts as a product. Liability is strict: an injured person doesn't have to prove anyone was negligent, only that the product was defective, that they suffered damage, and that the defect caused it. Covered damage includes death, personal injury (including psychological harm), damage to property, and loss or corruption of data.
The directive also makes these claims easier to bring against complex technology: courts can presume a product is defective in certain situations — for example where the product doesn't comply with mandatory product safety requirements, or where technical complexity makes proving the defect excessively difficult — and the defendant can rebut those presumptions.
Who it targets matters. Strict liability falls on the manufacturer and others in the supply chain, such as importers. AI Act providers are treated as manufacturers, and so is anyone who substantially modifies a product outside the manufacturer's control and then puts it back into service. A business that simply embeds a vendor's chatbot on its site is usually a deployer, not the manufacturer — so a claim under this directive would normally be aimed at the vendor.
Timing: Member States must transpose the directive into national law by December 9, 2026, and it applies to products placed on the market or put into service after that date. Until then, and for earlier products, the previous rules still apply.
Everything the product liability regime doesn't cover falls back on each Member State's own civil liability rules. That includes most claims against a website operator that uses an AI system rather than manufacturing it, and the kinds of loss the directive doesn't reach. These are generally fault-based: the claimant has to show you did something wrong, and the rules, standards of proof and remedies differ from country to country. This is the gap the AI Liability Directive was meant to harmonise, and it remains unharmonised.
AI Act obligations sit alongside both frameworks. Breaching Article 50 is a regulatory problem in itself — fines of up to €15 million or 3% of worldwide annual turnover under Article 99(4) (see penalties explained). But a documented failure to meet a legal obligation is also the kind of fact a claimant can point to when arguing a product was defective or that you were at fault.
How much weight it carries depends on the framework and the country — whether a transparency breach counts toward a presumption of defect, or as evidence of fault under national law, isn't settled yet. But an obvious, avoidable compliance gap doesn't help your side of any argument.
We don't assess your liability exposure. With three overlapping legal frameworks, national variations and questions about your role in the supply chain, that's a job for a qualified lawyer.
What we do is narrower and concrete: make sure you're not creating easily avoidable evidence against yourself. An undisclosed AI chatbot under Article 50(1) or AI-generated content published without a disclosure under Article 50(4) isn't just an Article 50 compliance gap — it's the kind of documented failure that strengthens a harmed party's case. Our scan finds those gaps on your public pages and tells you how to close them.
This page is general information, not legal advice. For your specific liability position, consult a qualified lawyer.
These are outcomes of fixing what the scan finds — not things the scan itself measures or scores.
No single fix eliminates liability risk. But addressing known transparency gaps removes one concrete piece of evidence a claimant could point to, and is the specific piece our scan checks.
The European Commission announced its withdrawal in its February 2025 work programme, citing no foreseeable agreement, and the proposal was formally withdrawn in October 2025. There is no dedicated EU AI liability law; the revised Product Liability Directive and national tort law cover the ground instead.
It depends on your role. The directive's strict liability falls on manufacturers and others in the supply chain — AI Act providers are treated as manufacturers. If you built the AI, or substantially modified it and put it into service, you may be in that position. If you only embed a vendor's chatbot, claims under the directive would normally be aimed at the vendor, and your own exposure would more likely come through national law.
Member States must transpose Directive (EU) 2024/2853 into national law by December 9, 2026. It applies to products placed on the market or put into service after that date; earlier products remain under the previous product liability rules.
No. Article50.io provides an automated technical assessment of your website's public-facing content against Article 50's transparency requirements, not legal advice. For a determination of your specific legal obligations or compliance strategy, consult a qualified lawyer.
Article50.io is an automated Article 50 transparency assessment platform that scans websites for potential EU AI Act transparency obligations and provides remediation guidance, implementation instructions, and compliance-ready disclosure language.
The free scan shows your single most severe finding in about 30 seconds — no signup, public pages only.
Automated technical guidance, not legal advice. Citations refer to Regulation (EU) 2024/1689.