Online service providers that use algorithms to determine how content on their platforms is displayed to users are not exempt from potential liability arising from that content – even where the source of the content is users themselves, the EU's highest court has ruled.
The Court of Justice of the EU (CJEU) addressed the issue in a ruling in which it considered the question of provider liability for content that is shared by users.
Under EU law, providers of 'information society services' are generally not liable for content on their platforms if they simply host the material that others publish. That position changes where hosting providers have actual knowledge of unlawful activity or information, in which case they must act expeditiously to remove or to disable access to the information to avoid liability.
In its ruling, the CJEU said the 'safe harbour' for hosting does not apply to an online service provider that stores information provided by users if it deploys an algorithm, "in its own interest or that of its service", to determine "under what conditions, how and in which order of priority that information is or is not broadcast as part of that service".
According to the CJEU, using an algorithm in this way constitutes exercising control over the information a user has shared, which lifts the provider out of the scope of the safe harbour for hosting.
Technology law expert Wesley Horion of Pinsent Masons said the CJEU's judgment significantly narrows the traditional understanding of "passive hosting" and is relevant to how many online platforms personalise services for users.
"This marks an important shift: content surfacing and amplification decisions, central to engagement-driven platform models, become directly relevant to liability, with the result that the same piece of unlawful content could trigger different outcomes depending on whether it is neutrally hosted or algorithmically promoted," Horion said.
Horion said the judgment is the latest by the EU court that puts constraints on the safe harbour protections for online platforms. He cited the CJEU's ruling in the case of Romanian website operator Russmedia Digital as another example in recent months, in which the court determined that online marketplaces are responsible for verifying personal information in adverts posted on their platforms. Horion said the erosion of safe harbour protections has potential implications for business models built on user engagement.
"Against that backdrop, and with the European Commission currently exploring further intervention through a potential Digital Fairness Act targeting addictive or manipulative designs, platform operators may need to anticipate a broader shift from neutral intermediaries towards more actively regulated actors, with end-to-end responsibility for how content is selected, surfaced and monetised," Horion said.
David Barker, who specialises in technology litigation and supports businesses across the technology, science and industry sector, said the AI age raises new questions about what intermediaries should and should not be deemed responsible for online.
"The EU's E-Commerce Directive enacted in 2000 marked a general international recognition that internet intermediaries should enjoy a high degree of protection from liability in relation to the use of their services by third parties," Barker said.
"Since then, we have had two very significant waves of technological advancement which have tested that proposition. The first was the introduction of Web 2.0 which enabled a dramatic increase in user-to-user interaction. If anyone were to look today at how the web looked back in the early 2000s they would find it extremely clunky: content on the internet was essentially static, not dynamic or interactive. Yet despite the quite dramatic changes which accompanied Web 2.0, the protections for intermediaries largely held up during this period of technological change."
"The advent of widespread use of AI represents a second major technological advancement and it is unsurprising that intermediary liability is again under the microscope. What needs to be resisted here, however, is the notion that intermediaries should assume liability for everything, in all circumstances, i.e. that the pendulum should now swing the other way. To take this approach would have far-reaching consequences not only for commerce but also for freedom of expression," Barker said.
Commercial contracts and AI expert Clare Francis, also of Pinsent Masons, said the direction of travel signalled by the courts and policymakers also has implications for retail businesses that rely on online marketplaces to reach customers.
"Erosion of the hosting defence isn't just a platform liability point – it shifts the commercial risk across the retail value chain," Francis said. "Any impact on the marketplace business model could result in an increase in seller fees and also raises questions around whether providers of the marketplaces can manage as many sellers, which in turn would impact the economics of the model."
"The ruling also means that marketplaces that curate and promote products will be considered more like actual retailers, meaning they need to consider a broader range of issues such as their exposure to product safety or other consumer claims, as well as brand liability," she said.
According to Francis, there are steps online platforms can consider taking to mitigate the increased risk of liability, such as re-negotiating seller terms, tightening their onboarding, and updating their compliance and audit programmes, to shift some risk down the chain.
"There will be limits to how far that works in practice and how much it affects the fundamental commercial proposition," Francis acknowledged.
For retailers, it will be important to understand the likelihood of being de-listed by platforms, what additional friction and compliance costs they might face if platforms impose new restrictions, and how platforms might change their approach to dealing with issues and enforcement of their terms and conditions, Francis said.
