Certified Curators

The proposal

Most regulatory answers to the power of social media platforms have depended on either public policy frameworks or self-regulation imposed by private players. Proposed measures have included breaking the platforms up, regulating their content from above, letting users carry their data from one to another, or limiting what may be done with that data. Up until recently, with the introduction of open protocols such as the AT Protocol or ActivityPub, users’ choice has been largely absent from the mainstream discussion. Users’ power has been limited to either abandoning these services or engaging in complex and burdensome content moderation procedures in case-by-case contestation.

The Certified Curators Model proposes a different approach: it starts with the users. It adapts the middleware idea originally designed by Professors Francis Fukuyama and Daphne Keller to the European Union context. The platforms keep performing the traditional function of data and identification hosts, but a new layer of companies and non-profits, the curators, is now invited to rank content that each user sees. Users can choose from a variety of curators in a competitive market of both for-profit, non-profit, private, and public algorithms. An independent EU agency certifies every curator before it launches and can withdraw its certificate, effectively excluding them from the market. Platforms can also propose their own Curator solutions, but these will be subject to self-preferencing limits.

Three Layers plus one Agency

01 · Platform Infrastructure

Platforms remain the places where people gather and interact. They keep the account information, the posts, the messages, and comments, and they keep improving the service as before. What they lose is the monopoly over ranking: they can still run a curator of their own, on the same terms as any other, but they stop being the sole gatekeepers of what people see. They are forced to open their data structures to Curators who explore for- and non-profit ways of content curation. In return, platforms negotiate under FRAND obligations, revenue-sharing deals with the new middleware Curator layer, benefiting from the success of these new entities.

02 · Curators

A new middleware market of entities connects to platforms through APIs and processes content where it sits, without transferring data. Each curator ranks by its own rules: a newspaper’s, a fan community’s, a fact-checker’s, or simply newest first.

Some curators are commercial and live on advertising or subscriptions. Others are non-profits, working in the public interest on fact-checking and disinformation. Others can be publicly run, for instance, by public service media or governments. All of them require certification and are bound by existing EU legislation on AI, competition, consumer protection and online expression.

A sublayer: content moderation

Illegal content is handled by a common content moderation sublayer, a necessary minimum that applies whichever curator a user picks, run by all the curators and the platforms’ teams together and fed by what each of them finds. Above that sublayer, curators compete on curation of lawful content.

03 · Users

Users stay on the platforms, with the communities they have built over a decade, and choose who curates for them. They gain agency through choice: a curator that serves its users badly loses them in a competitive market, whereas a curator that does not provide DSA guarantees can be de-certified and forced to leave the market.

The agency: a layer of certification

An independent EU agency certifies each curator before it launches, through a conformity assessment modeled on the AI Act, and goes on to check it afterward through the transparency reports and annual independent audits the DSA already provides for (Articles 15, 37 and 42). Certificates expire and have to be renewed. Rather than a new body, the role could go to the European Board for Digital Services, to a reinforced European Digital Media Observatory, or to the European Board for Media Services. Entities such as the European Digital Media Observatory could provide guidance in this process.

How is this sustainable?

Curation/targeted ranking is what pays for today’s platforms, so the model compensates them for losing this monopoly. Curators are forced to share part of their advertising or subscription revenue with the platforms on fair, reasonable and non-discriminatory (FRAND) terms, the kind of access terms EU law already sets in the Data Act (Article 8(1)) and the DMA (Article 6(5), (11) and (12)). The amounts are left to bargaining, with an independent third party to settle what the parties cannot, as in the framework the Court of Justice set out in Huawei v ZTE for standard-essential patents. Where a non-profit fact-checker does the work, for example, of identifying lawful but awful pieces of content to be shared among the Curators’ network, the money runs the other way: the platform shares part of its advertising profit with it.

A big discussion to be had is that advertising models for curators risk replicating the same problems of the past. Users may choose the most engaging curators, and curators may then compete on profiling, rebuilding the incentives of today’s platforms. Certification could favor contextual advertising and subscriptions over advertising based on profiling, and some limits already exist: the DSA prohibits advertising based on profiling with sensitive data or aimed at minors (Articles 26(3) and 28(2)), and since October 2025 political ads may be targeted only with the user’s explicit and separate consent.

Why it could actually work

Today the rules for online expression target companies that do both content hosting, moderation and curation, so every enforcement action risks jeopardizing the entire survival of the place where people gather. Under the Certified Curators Model, the rules would land on the curators, whose business is precisely curating and moderating, not hosting or authenticating. Nothing has to be broken up if new rules on algorithms are passed because there is now a competitive market of options, all fully aware of the rules and bound by a certification scheme that regularly attests to compliance.

On top of this, most of the conceptual legal framework already exists and there is no need to reinvent the wheel. Some examples include the notion of fair access and interoperability mandates in the Data Act and the DMA, conformity assessments in the AI Act, and transparency reports, audits, researchers’ access and out-of-court dispute settlement in the DSA. While such a regulatory push would be unlikely to succeed in the United States, it is not implausible to believe it would be legally possible to succeed in the EU.

We are not starting from scratch

Parts of the model are already running. On Bluesky’s AT Protocol, users have chosen among custom feeds built by third parties since 2023, and since March 2024 they can subscribe to independent moderation services, the so-called labelers, which anyone can run with Bluesky’s open-source tool, Ozone. European apps such as W, from Sweden, and mu, run by a Dutch foundation, now build on the same protocol, although both still read their timelines through Bluesky’s servers by default.

Unfortunately, it seems the largest platforms moved in a different direction. In 2023, X put most access to its API behind paid tiers and Reddit began charging for its own. Threads, Meta’s fediverse attempt, uses ActivityPub but does not offer fediverse sharing in Europe. There seems to be no will to abdicate curation control; instead, they continue to engage in practices that provide very limited choice to their users.

An open question: private data

A big question concerns public versus private access to data by curators. In this Certified Curators Model, curators would only read publicly available posts. It is up for discussion whether non-profit curators, for example, fact-checkers, should also read what is shared in private groups, for purposes of disinformation prevention. For it: much disinformation and illegal content circulates in closed groups, out of sight of any curator. Against it: the DSA prohibits the imposition of general monitoring obligations (Article 8) and the Court of Justice has repeatedly rejected the general and indiscriminate monitoring of communications. There are middle paths: access triggered by notices from trusted flaggers (Article 22 DSA), access on the model of vetted researchers (Article 40 DSA), or treating very large “private” groups as public spaces.