- A dispute about numbers
On 14 October 2026 the General Court delivered its judgement in Case T-134/24, Technius v Commission, dismissing in full the action brought by Technius, the operator of the adult platform Stripchat, against its designation as a very large online platform (VLOP) under the Digital Services Act (DSA). The judgement ruled on a dispute regarding a number of questions on the conditions for the designation of VLOPs and how these conditions should be assessed when they change over time.
The facts can be summarized as follows. On 15 February 2023, Technius published on its website an average monthly number of active recipients (AMAR) in the Union of about 31.8 million, below the threshold of 45 million set by the DSA. Its figure came from data of the analytics provider Similarweb: visitors using different devices were first deduplicated, then a bounce rate of 49% was deducted. The bounce rate reflects the proportion of ‘bounced visits’, i.e. cases where users had logged in to Stripchat but had not been exposed to any content disseminated on that platform (para. 5 of the judgement). The Commission accepted the Similarweb data, but not the deduction. In its view the metric of ‘unique visitors’ already filters out a large part of the bounced visits, so that the further discount was an artificial reduction. The Commission counted more than 62 million unique visitors per month and, by decision of 20 December 2023, designated Stripchat as a VLOP, together with two other adult platforms. A procedural twist deserves mention. On 27 May 2025 the Commission terminated the designation of Stripchat under Article 33(6) DSA and repealed the contested decision. The action went on nonetheless: the decision had produced effects from 22 April 2024 until 28 September 2025, and the Court held that Technius kept an interest in its annulment. The Court therefore ruled on a designation which no longer exists, and this is not irrelevant for what follows.
- The reasoning of the Court
The case was heard by the General Court and concerned an action brough by Stripchat to seek annulment of the designation decision adopted by the Commission. The Court rejected all the pleas, and four main findings can be summarised.
First, the Court noted that the statement of reasons provided by the Commission in its decision was adequate. The decision indicated clearly the shortcomings of the methodology of Technius and the material on which the Commission relied (para. 27 of the judgement). Since Technius had itself used the Similarweb data, and had not questioned their reliability when replying to the preliminary findings, the Commission could rely on them without explaining why they were reliable (para. 38 of the judgement). The Court then observed that the ‘unique visitors’ datum, as defined by Similarweb, already involves a filtering process which eliminates a significant number of bounced visits (para. 58 of the judgement). The Commission was therefore right to reject the further reduction of 49% (paras. 59 and 60 of the judgement). Nor was it obliged to calculate a different bounce rate, since Similarweb itself had confirmed that no technical means existed to state it with sufficient precision, and Technius had proposed no analysis during the administrative procedure (paras 61 and 62 of the judgement).
The second finding concerns the timing and retroactive assessment of the conditions for designation. According to Technius, corrected data obtained from Similarweb on 8 August 2024 would put the AMAR of Stripchat for the relevant period at about 37.3 million, hence under the threshold (para. 43 of the judgement). For the Court this is irrelevant: when the decision was adopted, neither the corrected data nor the new methodology belonged to any category of information listed in Article 33(4) DSA. Taking them into account retroactively would create legal uncertainty, since such data are subject to the volatility of the algorithms producing them, and the legality of an EU measure is assessed at the time of its adoption (paras 44 and 47 of the judgement). Moreover, the absence of a delegated act on methodology does not infringe legal certainty, which does not require that every rule ‘remove all doubt as to its interpretation’ (para. 85 of the judgement).
- A snapshot of a moving target
In my view, the most interesting passage is the one on the corrected data. The Court applied the classic approach of assessing the legality of an act at the date of its adoption. Yet its application to the DSA has a particular taste. AMAR is not a fact which can be verified as a turnover can. It is an estimate produced by the algorithm of a third party, and that algorithm changed: the 62 million of the Commission and the corrected 37 million come from the same provider of data. The designation stands, even if on the corrected basis the platform would have been below the threshold.
The system is not without a safety valve, since the Commission itself terminated the designation in May 2025. But the valve works ex post. In the meantime, for about seventeen months, the additional obligations on transparency, risk management and oversight were applicable. The concern of the Court is understandable, since a review of legality cannot chase every later revision of the data, and legal certainty would suffer. The price of this choice, however, is paid by the provider, who bears the volatility of a metric that nobody controls.
A conceptual remark may be added. The judgment describes a bounced visit both as a visit in which the user is not exposed to any content (para. 5 of the judgement) and as a visit in which the user leaves after viewing a single page (para. 51). The two are not the same thing: recalling recital 77 DSA, the Court itself notes that a recipient may be exposed to information simply by viewing it (para. 90 of the judgement). A visitor who looks at one page is therefore arguably an active recipient, and the deduction sat uneasily with the concept itself.
Regarding the concept of active recipient, a main point of contention concerned the fact that the absence of a delegated act made the designation not sufficiently foreseeable, to the detriment of legal certainty. The Court answered that the DSA gives sufficient guidance on the concept of active recipient, through Article 3(p) and recital 77 (paras 87 and 92 of the judgement), and that where a provision prescribes no method it is for the operators to choose a reliable one (paras 86 and 93 of the judgement). The Court also stresses that the concept must adapt to different services, including sites with adult content, where access may depend on proving the age of users (para. 87 of the judgement). Platforms, it adds, keep some discretion in choosing the method, being best placed to collect and analyse the data (para. 94 of the judgement). In this regard, it appears rather odd that the methodology on which designation rests is left to the very subjects to be designated, while the Commission keeps the power to reject their choice.
The point is not new. The Court builds on its own judgment in Zalando (T-348/23), noting that the mere possibility of a delegated act does not show that, in its absence, the regulation becomes ambiguous (para. 97 of the judgement). Zalando is under appeal, so Technius is best read as a consolidation, waiting for the Court of Justice.
- Conclusion
The Court gives a clear message in this judgement. The methodology used to count AMAR must be documented and defensible, deductions need a basis in the DSA, and the Commission will not clean the data in place of the provider. The points of contention in the judgement show that calculating the AMAR is not an easy task, it can raise doubts and create disagreement regarding the concrete definition of an active recipient. This is the case especially if one considers that services can be designed in different ways and thus providing for different channels of access to, and engagement with, the service. For the rest, the judgment can still be appealed on points of law within two months and ten days, and a delegated act under Article 33(3) would make much of this discussion obsolete.