Compliance · DSA

DSA for press websites: what publishers actually need.

Last updated: 6 mai 2026 · All claims sourced from Regulation (EU) 2022/2065 (the DSA) and Commission guidance.

What is the DSA, in one paragraph

The Digital Services Act (Regulation (EU) 2022/2065) is the EU law governing online intermediaries, including hosting platforms and any service that publishes user-generated content. It applies regardless of where the service is established, as long as it is offered to users in the EU (Art. 2). For press websites, this means comment sections, debate spaces and any reader contribution module fall under the regulation.

Does the DSA apply to your press website?

Yes, if you publish reader comments, host debates, or run any feature that lets users post content. The DSA distinguishes three tiers of obligations, with thresholds set in Article 19 and Article 33:

Online intermediaries, all services. Basic transparency and notice-and-action obligations.
Hosting services, the press websites running comments fall here. Article 16 obligation: notice-and-action mechanism for illegal content.
Online platforms, services that store and disseminate content to the public. Most press comment systems qualify. Articles 20-28 apply: complaint handling, transparency reports, terms and conditions clarity.
VLOPs / VLOSEs (Very Large Online Platforms / Search Engines), services with 45M+ active EU users. Most national press websites do not reach this threshold. Designation is published by the Commission (official list).

The four obligations that matter for press websites

1. Notice-and-action mechanism (Article 16)

You must provide an easy way for any user (or third party) to flag content as illegal. The mechanism must be electronic, sufficiently precise, and adequately substantiated. Acting on a notice triggers actual knowledge for the host (Art. 6.1 b), which means you can lose liability protection if you don't act diligently.

2. Statement of reasons (Article 17)

When you remove a comment, restrict its visibility, or suspend a user, you must provide the affected user with a clear and specific statement of reasons. The statement must include: which content is concerned, the territorial scope of the decision, the legal basis, the facts and circumstances relied upon, the use of automated means, and information on redress (Article 17, full text on EUR-Lex).

3. Transparency reports (Article 15 and Article 24)

All hosting services and online platforms must publish at least annually a public report on their content moderation activities. The report must include: orders received from authorities, notices submitted under Article 16, content moderation engaged at the provider's own initiative, complaints received via the internal complaint-handling system, and use of automated means. Online platforms (Article 24) face additional transparency obligations.

4. Terms and conditions (Article 14)

Your terms and conditions must include information on any restrictions you impose on user-generated content, including policies, procedures, measures and tools used for content moderation, including algorithmic decision-making and human review. The terms must be drafted in clear, simple, intelligible, user-friendly and unambiguous language, and made publicly available in an easy-to-find format.

What you need to ship by the deadlines

The DSA has been fully applicable since 17 February 2024 for all intermediary services (Article 93). VLOPs/VLOSEs were already obligated as of August 2023 (four months after their designation by the Commission).

Concretely, your press website should already have:

A working notice-and-action mechanism, easy to find from any comment-bearing page
A clear and specific statement-of-reasons template applied to every moderation decision
An annual transparency report, published in a publicly accessible location on your domain
Terms and conditions that explicitly describe your moderation rules and tools

How Logora handles Article 16 (notice-and-action)

Every comment, debate argument and consultation contribution carries a built-in reporting affordance, accessible to any reader (authenticated or not). When a user submits a notice, Logora :

Captures the content URL, the alleged violation type, and the optional user account context in a structured form.
Sends an acknowledgement of receipt to the reporter, with a reference identifier (Art. 16.4 of the DSA).
Routes the notice into the moderation queue with priority, so the moderation team or the algorithmic pipeline processes it without delay.
Records the timestamps for receipt, decision and notification, so the audit trail covers every reporting cycle end-to-end.

How Logora handles Article 17 (statement of reasons)

Every moderation decision generates a statement of reasons attached to the moderated content, automatically. Both parties are notified : the user whose content was actioned, and the user who reported it (when there was a notice).

Authors of rejected contributions see the decision and the reason category in their own profile history, so every interaction stays transparent. The redress link is included in the notification, in line with Article 17.3 (f). Six standard rejection categories are pre-mapped, each surfaced as an explicit user-facing reason :

Incivility — the contribution does not meet the basic editorial standards (Inzivilität).
Inappropriate language — obscene, vulgar, or coarse wording (Unangemessene Sprache).
Personal attack / hate — targeting an individual or a protected group (Persönlicher Angriff / Hass).
Incomprehensibility — the contribution is not understandable in its context (Unverständlichkeit).
Off-topic / advertising — the contribution is unrelated to the debate or promotional (Off-topic / Werbung).
Repetition / no substance — the contribution adds no argument (Wiederholung / Gehaltslosigkeit).

The wording of each user-facing notification is configurable per publisher, so it matches your editorial voice and your terms of service.

How the underlying pipeline works

Every contribution goes through three filters before reaching readers :

Blacklist check — the contribution is matched against the publisher's blacklist (publisher-defined explicit-rejection words) and a suspicious-words list. A hit on the blacklist auto-rejects with the corresponding statement of reasons.
Algorithmic moderation — an AI classifier scores the contribution. High-confidence clean contributions are auto-published. High-confidence violations are auto-rejected with a statement of reasons. Uncertain cases are queued for human review.
Human moderation — the moderation team reviews the queued contributions and makes the call with full editorial context. Decisions persist : a moderator who re-opens a previously decided contribution sees the prior decision and reason.

The reviewer interface includes keyboard shortcuts for speed (Ctrl + A accept, Ctrl + S skip, Ctrl + 1 to 6 for each rejection reason). Multi-select moderation lets the team process batches of related contributions in one action.

The infrastructure side

Transparency report exportable on demand from the administration space, formatted for the Commission's reporting database structure.
Editorial override : any automated decision can be reverted by your moderation team in real time, and the override is logged for the next audit cycle.
User bans : configurable durations (1 day, 1 week, 1 month, permanent), with the rationale captured in the user's profile and a Statement of Reasons sent on enforcement.
EU sovereignty : hosted on OVH, France. Data never leaves the EU. No Schrems II exposure on reader cookies or moderation logs.

Sources

Need to check whether your current setup is DSA-compliant? Book a 60-min call with our team and we'll review your moderation pipeline.