EU Product Liability: Version Every Product Claim

A product demo can remain technically flawless and still stop describing the product customers receive.

Imagine a launch film recorded on firmware 1.2. Voice control, response time and safety prompts were accurate on shoot day. Six months later, an update changes the workflow, but marketplace ads and distributor pages continue serving the old cut. Nothing is corrupted. No scene was fabricated. The asset has simply lost its product version.

For connected hardware, that is no longer a minor DAM problem. It is a product-truth problem.

Why the updated EU liability regime reaches the content library

Directive (EU) 2024/2853 must be transposed into national law by 9 December 2026 and applies to products placed on the market or put into service after 9 December 2026. The EU’s official summary says the framework updates product-liability rules for new technologies and gives businesses greater legal certainty.

The European Commission confirms that the new rules explicitly cover software, AI systems and product-related digital services. It also notes that, where the manufacturer is outside the EU, an economic operator in the Union must be available from whom an injured person can seek compensation; the provision can also reach products sold through online platforms. See the Commission’s 9 December 2024 explanation.

When assessing defectiveness, the directive includes the safety the public is entitled to expect, the presentation and characteristics of the product, reasonably foreseeable use and, where relevant, the effect of a product’s ability to learn or acquire new features after release. It also gives national courts a mechanism to order proportionate disclosure of relevant evidence in an accessible and understandable form.

The directive does not require brands to create a marketing-asset ledger, and an outdated video does not automatically establish a defective product. ONCE’s production inference is narrower: when software can change the behaviour of hardware after launch, a video that no longer identifies its product version becomes harder to approve, explain, update and retire responsibly.

Brands should obtain product-specific legal advice for the relevant member state, product category and supply-chain role. The method below is content governance, not a legal conclusion.

The file name is not enough

Most content libraries can tell you a video’s language, aspect ratio, channel and delivery date. Few can tell you which hardware revision, firmware, app build, cloud service or model produced the behaviour shown on screen.

That gap matters because a product claim is not limited to a voiceover sentence. A timed demonstration can imply speed. A sequence can imply that a safety step is unnecessary. A comparison shot can imply compatibility. Cropping a disclaimer or removing the page around a video can change the meaning without altering the master footage.

Connected products therefore need two kinds of version control: engineering version control for what the product does, and claim version control for what the market is being shown.

The ONCE five-field Claim Version Ledger

Start with a structured record attached to every final asset that makes or implies a material product claim.

Asset identity: asset ID, language, market, channel, aspect ratio and publication date;
Product identity: SKU, hardware revision, firmware, app build, cloud service and AI-model version where relevant;
Claim and presentation: the explicit words, implied performance, demonstrated conditions and any qualification viewers need;
Evidence and owner: the specification, test, certificate or approved product record, its date and version, plus the product, legal or safety owner who approved it;
Lifecycle: approval window, review date, change triggers and the person responsible for edit or withdrawal.

The point is not to create administrative theatre. It is to turn an exported video from an anonymous MP4 into an asset with a product identity, evidence basis and retirement condition. The relationship should also extend to captions, thumbnails, cut-downs, creator adaptations and translated voice tracks.

Four changes that should reopen the asset decision

1. Product identity changes

A component, connector, sensor, material or safety-related package instruction changes. Visual similarity is not enough to treat the new SKU as the same demonstrable product.

2. Software or AI behaviour changes

Firmware, an app, a related cloud service or a model update changes capability, output range, permissions, latency or a safety control. If the demonstrated behaviour no longer exists, retire the claim. If it remains available under narrower conditions, edit the qualification and check whether the scene itself still communicates the right expectation.

3. The evidence status changes

A certificate expires, a test method changes, a safety notice is issued or a claim source is superseded. Updating a subtitle may not be enough if the action on screen still implies the old conclusion.

4. Market or context changes

The same scene moves to another country, language, marketplace or creator account. Translation, removed captions and surrounding page copy can alter what an average viewer takes from the demonstration. Localisation is therefore a new claim review, not just a text service.

Keep, edit or retire

Keep: product identity, behaviour, evidence and use conditions are unchanged; only non-semantic formatting such as compression or a safe crop changes.
Edit: the core behaviour remains accurate but the asset needs new wording, qualification, safety information, market context or disclosure.
Retire or reshoot: the demonstrated version, material behaviour, operating condition or safety expectation has changed and cannot be corrected with text.

Distribution friction is not evidence that a video remains accurate. A distributor already having the file, an ad still performing or a reshoot being expensive are operational facts. They do not extend the truth window of a product claim.

Put version control into the next production purchase

A useful brief adds four inputs before the cameras roll: the sample’s exact hardware and software build; the evidence source behind every material claim; the owner who will alert the content team when the product changes; and the date or event that triggers review.

Delivery should include more than final exports. Ask for the Claim Version Ledger, editable masters, subtitle and voiceover source files, and a channel-version inventory. That makes a later correction an asset operation rather than an archaeological search across agencies, folders and retailer accounts.

A simple cost model is:

Version-governance cost = capture tagging + claim review + localisation review + periodic audit. Blind-reuse cost = channels × historical versions × discovery, withdrawal, editing and reshoot effort.

There is no honest universal industry average for these variables. The model is useful because it exposes when governance becomes economical: the more frequently a connected product changes and the more channels reuse its footage, the more expensive it becomes to discover too late which film belonged to which product.

For teams coordinating hardware, apps and content across China-based production, ONCE’s ecommerce product video production service can be briefed with version fields and claim owners from the start. A reusable asset is not one that never changes. It is one the brand can still prove belongs to the product.