Two important European timelines now sit close to each other, but they are not the same.
The revised Product Liability Directive applies to products - including AI systems - placed on the market or put into service from 9 December 2026.
The EU AI Act's high-risk obligations, following the Digital Omnibus changes, apply later:
- 2 December 2027 for most standalone high-risk systems listed in Annex III (for example systems used in biometrics, employment, education, access to essential services, law enforcement and similar areas).
- 2 August 2028 for high-risk AI systems that are safety components of, or are themselves, products covered by existing EU product-safety legislation listed in Annex I.
This gap of roughly one year (and longer for some product-embedded systems) has practical consequences. Liability exposure under the Product Liability Directive can begin before the full set of high-risk AI Act duties - such as the mandatory Quality Management System under Article 17, conformity assessment, and related high-risk requirements - becomes applicable to many systems.
Two different legal regimes
It is important to keep the regimes distinct.
The Product Liability Directive is about compensation when a defective product causes damage. It uses strict liability. From 9 December 2026 it clearly covers AI systems as products. A claim can succeed without proving that the manufacturer was negligent.
The EU AI Act is a regulatory framework. It imposes obligations on providers and deployers of AI systems, with the heaviest requirements falling on high-risk systems. Those high-risk requirements for Annex III systems apply from 2 December 2027 (and later still for many Annex I cases).
One regime is about private liability and compensation. The other is about regulatory compliance and market access conditions. They interact, but they do not share the same start date for the most demanding AI obligations.
Why the earlier liability date matters
From 9 December 2026, an organisation that places an AI system on the EU market can face a product liability claim if that system is defective and causes covered damage. In complex AI cases, claimants may benefit from court-ordered disclosure and from presumptions of defect and causation.
One of those presumptions is particularly important: defectiveness may be presumed where the product does not comply with mandatory product safety requirements laid down in Union or national law that are intended to protect against the type of damage suffered. As AI-related safety requirements become mandatory, failure to meet them can feed directly into a liability claim, shifting the burden onto the manufacturer to rebut the presumption.
In practical terms, this means that the quality of design control, risk management, technical documentation, testing, update processes and post-market monitoring can matter for liability defence from December 2026 - even if the full high-risk regulatory machinery of the AI Act is not yet mandatory for that system until December 2027 or August 2028.
What is not automatically pulled forward
The Product Liability Directive does not formally move the AI Act's application dates. High-risk obligations under the AI Act still apply according to the Omnibus timeline. Organisations are not legally required to complete conformity assessment or operate a full Article 17 Quality Management System for Annex III systems solely because of the Product Liability Directive's start date.
What the earlier liability date does is change the risk and evidence picture. If a claim arises after 9 December 2026, the organisation will need to explain and evidence the safety of the system under the new liability rules. Waiting until the 2027 or 2028 regulatory deadlines to build documentation, ownership, testing records and monitoring capability leaves a period in which liability exposure exists without the full regulatory compliance framework necessarily being in place.
Other AI Act dates still matter
Not everything in the AI Act was postponed. Transparency obligations, rules on prohibited practices, general-purpose AI obligations and other provisions have their own timelines, many of which are already in effect or apply during 2026. Where those rules constitute mandatory requirements relevant to safety, they can also become relevant to the assessment of defectiveness under the Product Liability Directive.
The practical implication for planning
Organisations placing AI systems on the EU market need to plan for two clocks:
- Liability clock: 9 December 2026 - strict liability regime applies to AI products placed on the market or put into service from that date.
- High-risk regulatory clock: 2 December 2027 (Annex III) and 2 August 2028 (many Annex I systems) - full high-risk AI Act obligations apply.
The first clock arrives earlier. Preparing technical documentation, design and risk evidence, update and monitoring records, and clear accountability is therefore not only preparation for 2027/2028 regulatory compliance. It is also preparation for liability exposure that begins in December 2026.