Part I — The Manufactured Crowd. Bots, Artificial Engagement and the Illusion of Public Opinion

Are We Still Talking To Humans?

Part I — The Manufactured Crowd

Bots, artificial engagement and the illusion of public opinion

You open Instagram, LinkedIn, TikTok or X.

A post has thousands of likes. Hundreds of people have commented. The same opinion appears again and again. Some comments are angry, others supportive. People argue with each other.

Without consciously deciding to do so, we register something:

Apparently, a lot of people think this.

That assumption matters.

Likes, comments, followers and shares are not just numbers. They are social signals. They help us estimate what other people find important, credible, outrageous or normal.

But there is a problem.

We increasingly cannot be certain how many independent human beings those numbers actually represent.

When 100 voices are not 100 people

Bots have existed for years. At their simplest, they are software programs that perform actions automatically: posting, following, liking, sharing or commenting.

One bot is not particularly interesting.

Thousands of automated or centrally coordinated accounts are.

If one organisation controls 100 accounts and all of them participate in the same discussion, we may perceive 100 different voices.

The interface shows 100 participants.

The reality behind the screen may contain only one organising actor.

That is where bots become more than spam.

They can create the appearance of popularity, outrage, disagreement or consensus. They can amplify a subject until real people start noticing it — and then those real people begin reacting as well.

Artificial activity can therefore generate genuine human activity.

At that point, separating the two becomes considerably harder.

The information does not even have to be false

Much of the public debate about online manipulation concentrates on disinformation: deliberately false or misleading information.

But manipulation can be more subtle.

Imagine a perfectly factual article that is artificially shared thousands of times by coordinated accounts.

The article has not become false.

Its apparent social importance has been manipulated.

That distinction matters because people do not judge information in isolation. We look at what others are reading, sharing and saying.

A minority opinion can be made to look widespread.

A controversy can be made to look enormous.

Anger can be made to look universal.

And repetition can create the impression that an issue is suddenly everywhere.

Nobody needs to put an opinion directly into our heads.

Changing the environment in which we form that opinion may already be influential.

This is not hypothetical

The European External Action Service (EEAS), the diplomatic service of the European Union, investigates what it calls Foreign Information Manipulation and Interference (FIMI).

Its 2026 threat report examined organised attempts to manipulate information environments. The report describes networks of channels, websites and social-media accounts used to produce and amplify influence operations.

Political influence is one documented application, but artificial engagement is not exclusively political.

Commercial actors can want popularity. Fraudsters need credibility. Marketers want reach. Campaign organisations want attention. And real users can organise themselves into so-called engagement pods, agreeing to like and comment on one another's posts to increase visibility.

Even LinkedIn explicitly prohibits fake accounts, bots and fake engagement intended to manipulate its content algorithms.

In March 2026, LinkedIn went further. It announced additional measures against automated comments and engagement pods, stating that LinkedIn should be a platform for real people, real jobs and real conversations.

That is an interesting choice of words.

Because apparently, even on a professional network, whether the person and interaction are real can no longer simply be assumed.

What does a like actually mean?

This changes the way we should look at familiar social-media statistics.

1,000 likes do not necessarily equal 1,000 independent human judgements.

500 similar comments do not constitute an opinion poll.

A trending subject does not automatically represent spontaneous public interest.

But the opposite is equally important.

Ten likes do not mean that only ten people found something valuable. Algorithms, network position, timing and recommendation systems also determine what becomes visible.

Social-media metrics look precise.

What they measure is considerably less precise than the number suggests.

That should make us cautious — not paranoid.

Not every strange account is a bot. Not every popular opinion is manufactured. And an opinion we disagree with should never be dismissed as artificial without evidence.

Awareness means understanding that manipulation is possible without imagining it everywhere.

But there is another problem

So far, we have assumed that bots are relatively easy to understand.

A machine performs actions that normally belong to a person.

But generative Artificial Intelligence changes something more fundamental.

AI can now create the photograph.

It can write the biography.

It can produce the posts.

It can write the comments.

It can change tone, language and personality.

And suddenly the question is no longer only whether the crowd is real.

It becomes much more personal:

Is the person I am talking to a person at all?

💠

The factual foundation for Part I comes from the European External Action Service (EEAS) and its 2026 report on Foreign Information Manipulation and Interference (FIMI). � A particularly well-documented historical example is the US case concerning the Internet Research Agency: the indictment describes hundreds of fictitious US personas and accounts designed to appear American. � LinkedIn currently prohibits fake accounts, bots and fake engagement and says it limits engagement pods and automated comments. �

Part II: Who Are You? — Synthetic identities and the disappearance of certainty online.

Part II — Who Are You? Synthetic Identities and the Disappearance of Certainty Online

Are We Still Talking to Humans?

Part II — Who Are You?

Synthetic identities and the disappearance of certainty online

There is a photograph.

A friendly face. A name. An age seems roughly apparent. There are holiday photographs, opinions, interests and a short biography.

The account occasionally responds to you.

What do you actually know about the person behind it?

Less than ever before.

For most of the history of social media, we have operated with an unspoken assumption:

There is a human being behind the profile.

Perhaps that person uses an old photograph. Perhaps the job title is exaggerated. Perhaps the carefully selected holiday photographs present a rather flattering version of reality.

But there is a person.

Generative AI is making even that basic assumption increasingly unreliable.

A person can now be manufactured

A realistic human face no longer requires a human model.

Neither does a voice.

A biography can be generated. Photographs can be created or altered. Posts can be written automatically. Conversations can be assisted or generated by AI.

And the identity can be changed.

The same operator could present one account as a young professional, another as a retired grandmother and another as someone of a completely different age, sex or nationality.

The relevant concept here is a synthetic persona: a digitally constructed identity that can combine generated images, invented personal information, AI-generated content and automated or human-controlled interaction.

Not every synthetic element makes an account deceptive.

Someone may use an avatar for privacy. A real person may use AI to improve a photograph or write a post. Millions of people legitimately use AI as a tool.

The crucial distinction is not:

Was AI used?

It is:

Is someone deliberately creating a false impression about who — or what — is communicating?

The problem becomes different when the user is a child

For adults, uncertainty about identity can lead to manipulation, fraud or deception.

For children, the stakes can be considerably higher.

A 13-year-old may believe another account belongs to a 14-year-old.

The profile photograph looks right. The language sounds right. They follow the same creators, know the same games and use the same expressions.

But none of those signals now proves that another 14-year-old is sitting at the other end.

An adult pretending to be a child online is not a problem created by AI. Grooming and false identities existed long before generative technology.

What AI changes is the available toolkit.

Convincing images, text and potentially voices can be generated or manipulated. Maintaining a plausible identity can become easier. And the familiar signals people once used to judge who someone might be become less reliable.

The European Commission already recognises grooming and unwanted contact from strangers as significant online risks to minors under the Digital Services Act (DSA).

But age and identity are two different questions.

Knowing that a user is over 18 does not tell us who that user is.

And knowing that an account claims to be 14 does not prove that the person operating it is 14.

Why not simply verify everybody?

At first sight the solution appears obvious:

Make everyone prove their identity.

But that creates another problem.

There are legitimate reasons to use the internet without publishing one's legal identity. Whistleblowers, abuse victims, political dissidents and ordinary citizens may have very good reasons not to make their identity public.

And requiring billions of people to hand identity documents to social-media companies creates its own privacy and security risks.

Europe is already wrestling with this tension.

The European Commission's approach to age verification is deliberately designed around privacy-preserving proof of age. A person could prove, for example, that they are over 18 without revealing their exact age, name or other unnecessary personal information.

That is an important distinction.

We do not necessarily need to know who someone is in order to establish what they are entitled to prove.

Perhaps the same kind of thinking will eventually become relevant to human authenticity.

Not necessarily:

Show me your passport.

But perhaps:

This platform has verified that a natural person controls this account.

That would be a very different proposition.

Platforms are already moving

LinkedIn is particularly interesting because professional identity is central to the platform.

In March 2026, LinkedIn reported that more than 100 million members had added at least one verification to their profile. It has also made workplace verification mandatory for members adding recruiter job titles.

Instagram and Facebook owner Meta takes a somewhat different approach, focusing partly on identifying and labelling AI-generated content.

TikTok likewise requires realistic AI-generated images, audio and video to be labelled in specified circumstances and can automatically attach an AI-generated label when it detects recognised technical credentials.

These are meaningful developments.

But notice what they primarily address:

the content.

Was this photograph generated?

Was this video manipulated?

Was this audio synthetic?

Our emerging problem goes one level deeper.

What about the identity producing all that content?

Content authenticity is not identity authenticity

A synthetic profile could contain a mixture of genuine and artificial material.

A real photograph could be stolen.

A generated photograph could represent a fictional person.

A human could write some posts while AI writes others.

An operator could personally answer important messages while automating routine conversations.

The account might therefore not be purely “human” or purely “AI”.

That binary distinction may already be becoming obsolete.

The more useful questions may be:

Who controls this identity?

Does it represent a real individual?

How much of its interaction is automated?

And is the person interacting with it entitled to know?

Those questions become especially important when trust is involved: friendship, recruitment, dating, financial advice, political discussion — and contact with children.

We used to ask whether the information was real

That question remains important.

But social media are moving into territory in which another question precedes it.

Before asking whether we can trust what somebody tells us, we may increasingly need to ask:

Who is telling us?

And sometimes:

Is there a “who” at all?

The internet has spent decades teaching us not to believe everything we read.

Generative AI presents a different challenge.

We may no longer know who we are reading at all.

💠

For this second part, the European Commission's DSA guidance specifically identifies grooming among the online risks platforms should address for minors. � The EU is simultaneously developing privacy-preserving age verification: its proposed approach can establish that someone meets an age threshold without disclosing their precise age or identity to the service. � TikTok already requires labels for certain realistic AI-generated content, including substantial face swaps, and supports automatic detection through Content Credentials. � Meta similarly labels detected AI-generated material on Facebook and Instagram. �

digital-strategy.ec.europa.eu

digital-strategy.ec.europa.eu +1

TikTok Ondersteuning

Facebook +1

Part III: The Right to Know — When transparency becomes a responsibility.

Part III — The Right to Know. When Transparency Becomes a Responsibility

Part III — The Right to Know

When transparency becomes a responsibility

Suppose an AI system answers a question on a company's website.

It says clearly: You are chatting with our virtual assistant.

No problem.

You know what you are dealing with.

Now imagine exactly the same technology operating through an Instagram account with a human name, a realistic face, a biography, photographs and opinions.

It comments on your posts.

You answer.

Over time, perhaps you begin to trust it.

The technology may be similar.

Your understanding of the interaction is completely different.

That difference is where transparency becomes important.

Europe has already recognised the principle

The European Union's Artificial Intelligence Act (AI Act) contains transparency obligations that became applicable on 2 August 2026.

Article 50 addresses several situations in which people should be informed about the involvement of AI.

Providers of certain AI systems must ensure that people are informed when they are interacting directly with an AI system, unless this is already obvious from the circumstances.

Providers of generative AI must also make certain AI-generated or manipulated output detectable in machine-readable form. Additional disclosure requirements apply to areas including deepfakes and some AI-generated text concerning matters of public interest.

The underlying principle is significant:

People should not unknowingly mistake artificial interaction or synthetic material for something human or authentic.

But social media are rapidly making the boundaries more complicated.

What exactly should be labelled?

Consider four accounts.

Account A belongs to a real woman who writes every word herself.

Account B belongs to a real woman who uses AI to help write her posts.

Account C belongs to a real person but uses a generated face, fictional biography and automated AI responses.

Account D represents no real individual identity at all and operates as a synthetic persona.

Simply attaching “AI-generated” to individual photographs does not explain those differences.

And yet those differences matter enormously to someone deciding whether to trust, employ, befriend or simply believe the account.

This suggests that content transparency and identity transparency are not the same thing.

Our regulatory frameworks are beginning to address the first.

The second deserves much more attention.

A right to know you are not talking to a human?

That raises an uncomfortable question.

Should a person have a right to know when an apparently human interaction is substantially automated?

The AI Act already embraces that principle in certain direct human-AI interactions.

But social-media identities can blur the categories.

What if a human controls the account but AI produces 90 per cent of its conversations?

What if ten people operate one persona?

What if one person operates fifty personas?

What if the photographs and biography are synthetic but every conversation is manually written by a human?

And what if an AI agent increasingly operates an identity independently?

The question “Is this a bot?” becomes surprisingly inadequate.

What matters is whether the user has been given a materially false impression about the nature of the entity with which they are interacting.

Platforms know more than their users do

Platforms occupy a unique position.

They can observe account histories, login patterns, automation signals, device behaviour, network structures and activity that ordinary users cannot see.

LinkedIn says it detects suspicious behaviour using technology and human review and can limit automated comments and inauthentic engagement. It is simultaneously expanding identity and workplace verification.

Meta uses technical standards to identify some AI-generated content.

TikTok can automatically recognise certain AI-generated material carrying Content Credentials.

The ordinary user has none of those capabilities.

That produces an important question of responsibility.

If a platform is better able than its users to identify artificial behaviour, how much of the burden should reasonably remain with the individual user?

“Be careful online” cannot be the entire answer.

But verification has a price

There is an obvious danger in demanding universal identity verification.

A social network on which every participant must disclose a government identity would solve one problem by creating several others.

Privacy matters.

Anonymity matters.

Freedom of expression matters.

People sometimes need pseudonyms.

And centralising identity information creates valuable databases that themselves require protection.

The objective should therefore not automatically be to abolish anonymity.

Perhaps the more useful distinction is between anonymity and deception.

A user might legitimately say:

I do not wish to tell you my legal name.

That is different from saying:

I am a 14-year-old girl, while deliberately constructing an entirely fictitious identity to make another person believe it.

Protecting anonymity does not necessarily require protecting deceptive impersonation.

That distinction deserves serious legal attention.

Awareness cannot carry the entire burden

For now, users can do something.

We can question unusual accounts. We can look beyond profile photographs. We can stop treating follower numbers and comments as straightforward measures of human opinion. We can be cautious when strangers seek personal contact.

But there is a limit to what individual vigilance can achieve.

A human being cannot reliably detect every AI-generated face.

A child cannot reasonably be expected to conduct digital forensics before trusting another apparent child.

And ordinary users cannot see the technical information available to the platforms hosting these interactions.

That is why this issue cannot end with digital literacy.

It becomes a question of platform design, verification, transparency and ultimately regulation.

The social contract of social media is changing

Social media were built around a simple idea:

people connecting with people.

Artificial Intelligence does not make that idea obsolete.

It does make it less safe to assume.

There is nothing inherently wrong with using AI to create, write, translate or communicate. AI can expand human creativity and make knowledge accessible in extraordinary ways.

The line becomes different when technology is deliberately used to manufacture human identity, human support or human consensus while the recipient is encouraged to believe it is genuine.

Perhaps that is the principle around which the next stage of regulation should develop:

Not compulsory exposure of everyone's identity.

Not suspicion of everything artificial.

But meaningful transparency when technology is being used to imitate the presence, identity or independent judgement of another human being.

Because in a digital society, knowing what we are looking at is becoming inseparable from knowing who — or what — is looking back.

Awareness matters.

But awareness cannot carry the entire burden.

JAS — Aware Solutions

💠

The legal backbone of Part III is Article 50 of the EU AI Act, applicable since 2 August 2026. The European Commission says it addresses risks including impersonation and manipulation and requires transparency in specified direct AI interactions as well as machine-readable marking of AI-generated or manipulated material. � Meta has meanwhile joined the EU code concerning transparency of AI-generated content. � LinkedIn's 2026 measures are particularly relevant to the identity question because it is simultaneously fighting automation and expanding profile and workplace verification. �

The Legal Work I Want to Do: Preserving Human Judgment in a Digital World

Digital platforms increasingly determine what people may say, see and share. They remove content, suspend accounts, restrict visibility and make decisions that can affect a person’s reputation, livelihood and ability to participate in public debate.

These decisions are necessary in some cases. Platforms have legitimate responsibilities regarding illegal content, threats, abuse and public safety.

But content moderation also raises difficult questions.

Who decides what constitutes disinformation? When does harmful expression become illegal content? When is removing content justified, and when does it disproportionately restrict freedom of expression? Were the platform’s rules clear, accessible and consistently applied?

These questions lie at the intersection of law, technology and fundamental rights. That is where I want to work.

A field that genuinely interests me

I am a Dutch-qualified legal professional with an LL.M. in International Law and Legal Studies. My strengths lie in legal research, source verification, contract analysis and clear legal writing.

In recent years, my independent research and publication work has increasingly focused on digital governance, privacy, consumer protection, artificial intelligence and the relationship between technology and individual rights.

My professional path has not been conventional. However, it has taught me to work independently, examine sources critically and approach complex questions without relying on ready-made conclusions.

I know how to identify the decisive issue, separate facts from assumptions, compare competing arguments and formulate a reasoned conclusion.

Disputes concerning content moderation bring these skills together in a field that I find genuinely compelling.

The importance of independent assessment

A platform dispute cannot always be resolved by mechanically applying a rule.

An assessor must first determine what happened. Which decision is being challenged? What reason did the platform provide? Which provision of its terms and conditions applies? What arguments and evidence has the user submitted?

The next question is whether the platform applied its rules correctly, consistently and proportionately. Applicable law and fundamental rights may also need to be considered.

Personal agreement with the disputed content should not determine the outcome.

Impartiality does not mean that an assessor has no personal values. It means recognising those values and preventing them from replacing the relevant legal and procedural standards.

There is an essential difference between content that I personally dislike and content that violates an applicable rule or legal standard. A fair assessment must preserve that distinction.

A structured approach to complex cases

When examining a case, I would begin by separating its different layers.

First, I would establish the relevant facts and identify the precise platform decision under review.

Second, I would examine the platform’s reasons and the applicable terms and conditions.

Third, I would assess the arguments and evidence submitted by the parties.

Fourth, I would identify the relevant provisions of the Digital Services Act and any other applicable legal principles.

Finally, I would formulate a clear and transparent conclusion explaining not only the outcome, but how that outcome was reached.

I would not begin with an intuitive conclusion and then search for arguments to support it.

Large case files become manageable when they are reduced to their essential components: the contested decision, the relevant facts, the applicable rule, the arguments of the parties, missing or contradictory information and the questions that must ultimately be answered.

The purpose of structure is not to oversimplify a case. It is to prevent the decisive issue from disappearing beneath the volume of information.

Clear legal reasoning

A decision must be understandable to the person affected by it.

Legal writing should identify the issue, explain the relevant standard, address the central arguments and show how the conclusion follows from the facts.

It should not be written to demonstrate how complicated the law can be.

This is particularly important in disputes between individuals and large digital platforms. A person should be able to understand why a decision was made, even if that person disagrees with the outcome.

Legal complexity should never become an excuse for unclear reasoning.

AI as a tool, not a decision-maker

I recognise the value of technology and artificial intelligence as professional tools.

AI can help organise information, compare documents, identify questions and improve efficiency. Subject to proper authorisation and safeguards, it may support legal research, case structuring and the review of draft texts.

However, I am strongly opposed to allowing technology to replace human responsibility in decisions affecting people’s rights.

Confidential or identifiable case information should never be entered into a public AI system without explicit authorisation. Any professional use of AI must comply with privacy, confidentiality and data-security requirements.

More fundamentally, an assessor must remain responsible for the substance of the decision.

AI output is not a legal authority. It may contain errors, overlook context or reproduce hidden assumptions. Facts, sources, inferences and conclusions must therefore be independently verified.

Efficiency is valuable. Human judgment, accountability and fairness are indispensable.

The human reality behind digital disputes

Platform cases are not merely administrative units to be processed.

Behind every file is a person who may have lost access to an account, had content removed or seen an important means of communication restricted. The platform may also have legitimate reasons for intervening, including protecting other users or complying with the law.

Both sides must be taken seriously.

That does not mean that every complaint is justified or that every platform decision is wrong. It means that the dispute deserves a careful assessment based on facts, applicable rules and transparent reasoning.

Out-of-court dispute settlement under the Digital Services Act creates an important space between an internal platform appeal and proceedings before a court. Its credibility will depend substantially on the independence, competence and care of the people conducting those assessments.

A field in which I want to develop

I do not regard this merely as an interesting temporary assignment.

I can genuinely see myself developing in this field over the next ten to fifteen years: first by becoming a reliable and knowledgeable assessor and later, perhaps, by contributing to the review of complex cases, quality standards, training, DSA policy or responsible AI governance.

Digital disputes will not disappear. As platforms and automated systems become more influential, questions about accountability, transparency, proportionality and fundamental rights will become even more important.

My professional background may be unconventional, but my direction is becoming increasingly clear.

I want to work where legal rules, platform governance and individual rights intersect.

And I want to help ensure that, in an increasingly automated world, meaningful human judgment does not disappear.

Trust Your Feed? A Legal Stress Test for Europe’s New Social Platform

Earlier this year, W Social was announced as a new European social media platform: built, governed and hosted in Europe, with “human verification, free speech and data privacy at its core.”

The initiative was introduced at the beginning of the World Economic Forum week in Davos. That timing does not, by itself, establish any formal affiliation with the WEF. It does, however, make transparency about ownership, financing and governance even more important.

The promise is attractive. Europe needs credible alternatives to platforms whose business models reward outrage, polarisation and maximum engagement.

But trust is not created by branding.

It has to be designed into the platform’s legal, technical and institutional architecture.

Free speech — defined by whom?

W Social presents freedom of expression as one of its foundations, while also identifying systemic disinformation as a threat to public trust and democratic decision-making.

Both concerns are legitimate. Their combination nevertheless raises the central question:

Who decides what constitutes disinformation, and according to which publicly accessible standard?

Illegal content can be assessed against applicable law. “Disinformation” is a far less precise category. Information may be incorrect, disputed, incomplete, unpopular or politically inconvenient without being unlawful.

A platform committed to freedom of expression therefore needs more than a general promise. It needs clear distinctions between:

• illegal content;
• content prohibited under its contractual rules;
• demonstrably false factual claims;
• disputed interpretations;
• opinion, satire and political criticism.

Without those distinctions, “trust” risks becoming an editorial label applied by an unidentified authority.

Human verification and data privacy

Several responses to the announcement questioned whether human verification can genuinely coexist with privacy and free expression.

That concern is justified, although human verification does not necessarily have to mean public identification. A platform could verify that a unique human being exists while still allowing that person to operate under a pseudonym.

The decisive questions are therefore technical and legal:

• What information must users provide?
• Will government-issued identification be required?
• Who performs the verification?
• Is identifying information retained after verification?
• Can verification data be connected to a user’s posts and network?
• Will biometric data be processed?
• Can users participate pseudonymously?
• How are people protected against profiling, data breaches and compelled disclosure?

Under the GDPR, collecting personal data because it might later prove useful is not enough. Verification must have a defined purpose, a lawful basis and a proportionate design. Data minimisation and storage limitation must be built into the system.

“Verified human” and “identifiable speaker” are not synonymous. That distinction should be explicit.

Governance is not a footnote

One commenter asked whether W Social would include collaborative democracy or meaningful citizen participation in its governance.

That may be the most important question of all.

Who owns the platform? Who appoints its leadership? Who finances its development? Who may change the rules? Who supervises the moderators? What happens when commercial interests conflict with freedom of expression?

A platform cannot credibly ask users to “trust their feed” while leaving its own power structure opaque.

User participation alone is not sufficient, but meaningful institutional safeguards could include:

• published governance arrangements;
• transparent ownership and funding;
• independent supervisory members;
• public content-moderation standards;
• structured consultation before material rule changes;
• transparency reports;
• external audits;
• protection against interference by investors, advertisers and political actors.

The relevant question is not whether the people in charge have good intentions. It is whether the governance model remains trustworthy when interests begin to conflict.

The business model shapes the public conversation

Another response correctly connected platform governance to the underlying business model.

If revenue depends primarily on advertising, surveillance and engagement, the platform may reproduce the very incentives it promises to escape. Content that triggers anger, fear or conflict often keeps users active longer.

A different interface does not create a different ecosystem if the economic logic remains unchanged.

W Social should therefore explain:

• how the platform will generate revenue;
• whether personal data will be used for behavioural advertising;
• whether engagement determines visibility;
• whether users can choose a chronological or unprofiled feed;
• and whether commercial partners can influence reach or moderation.

Content governance cannot be separated from revenue architecture.

Open infrastructure or another closed platform?

Technical contributors also asked whether W Social is based on open protocols such as AT Protocol or ActivityPub, whether it will be federated and whether users will be able to communicate across networks.

These are not merely technical questions.

Interoperability affects competition, user autonomy and the practical ability to leave a platform without losing one’s entire digital network. Portability should mean more than downloading an archive that cannot be used elsewhere.

If W Social depends on technology developed by another social network, it should also clarify which components are genuinely open source, which remain under third-party control and what dependencies this creates at European scale.

A European server location alone does not establish European digital sovereignty.

Moderation requires due process

Every social platform eventually restricts content or accounts. The true test of its commitment to freedom of expression begins at that moment.

The EU Digital Services Act already provides an important procedural framework. Platforms must communicate their restrictions clearly, establish internal complaint mechanisms where applicable and allow users access to certified out-of-court dispute settlement.

A credible platform should therefore be able to answer:

• Are moderation decisions made by humans, automated systems or both?
• Will users receive a specific statement of reasons?
• Can they challenge the evidence and interpretation used?
• Who decides the internal appeal?
• Is the reviewer independent from the original decision-maker?
• Can comparable cases produce comparable outcomes?
• Is external dispute settlement genuinely accessible?
• Will important decisions be published in anonymised form?

An appeal process is not a customer-service feature. It is part of the platform’s rule-of-law infrastructure.

Trust must be contestable

W Social’s announcement has generated precisely the questions a serious European platform should welcome.

The critical responses do not prove that the initiative will fail. Nor does its introduction during the WEF week prove who controls or finances it.

They do show that its central claims remain to be demonstrated.

A trustworthy platform does not merely promise correct decisions. It defines its powers, limits data collection, explains its reasoning and provides meaningful redress when it gets things wrong.

Europe does not simply need another social network.

It needs digital institutions in which power is visible, decisions are reviewable and fundamental rights do not depend on corporate goodwill.

Only then does “Trust your feed” become more than a slogan.

💠💠

Enkele belangrijke afkortingen

DSA — Digital Services Act

Europese regels voor onlineplatforms.

GDPR / AVG — General Data Protection Regulation / Algemene verordening gegevensbescherming

Europese privacywetgeving.

ADR — Alternative Dispute Resolution

Alternatieve/buitengerechtelijke geschillenbeslechting.

ODS — Out-of-Court Dispute Settlement

Buitengerechtelijke behandeling van platformgeschillen onder de DSA.

DSC — Digital Services Coordinator

Nationale toezichthouder op de DSA.

VLOP — Very Large Online Platform

Zeer groot platform, zoals Instagram of TikTok.

VLOSE — Very Large Online Search Engine

Zeer grote onlinezoekmachine.

T&C — Terms and Conditions

Algemene gebruiksvoorwaarden van het platform.

AT Protocol — Authenticated Transfer Protocol

Het decentrale netwerkprotocol achter Bluesky.

ActivityPub

Open protocol waarmee platforms binnen het fediverse met elkaar kunnen communiceren.

100 % Remote paradox

100% remote: hoe locatie-onafhankelijk is een arbeidsrelatie juridisch? De juridische grenzen van werken zonder vaste werkplek 

“100% remote.” “Work from anywhere.” “Remote – Europe.”

Voor werkzoekenden klinken deze termen alsof locatie nauwelijks nog relevant is. Voor steeds meer werkzaamheden klopt dat technisch ook. Een laptop, internetverbinding en toegang tot de systemen van de werkgever kunnen voldoende zijn om hetzelfde werk vanuit Rotterdam, Parijs of Rome te verrichten.

Juridisch ligt dat anders.

Zodra werknemer en werkgever zich niet in hetzelfde land bevinden, komen vragen op over arbeidsrecht, sociale zekerheid, belasting en werkgeversverplichtingen. Een functie kan dus technisch volledig locatie-onafhankelijk zijn, terwijl de arbeidsrelatie juridisch sterk aan landsgrenzen gebonden blijft.

Dat maakt één vraag bij internationale recruitment steeds belangrijker:

Wat betekent “100% remote” eigenlijk?

Remote is niet hetzelfde als work from anywhere 

De term remote zegt in de eerste plaats iets over de fysieke werkplek: de werknemer hoeft niet structureel naar een kantoor van de werkgever.

Daaruit volgt niet automatisch dat die werknemer vanuit ieder willekeurig land mag werken.

Een werkgever kan goede redenen hebben om een remote functie alleen open te stellen voor werknemers die in bepaalde landen wonen of werken. Zodra iemand structureel vanuit een ander land werkt, kunnen daar registratieverplichtingen, sociale premies, fiscale gevolgen en dwingendrechtelijke arbeidsregels ontstaan.

Daarom kan een vacature tegelijkertijd 100% remote zijn én geografisch beperkt.

Het probleem ontstaat wanneer die beperking pas ná de sollicitatie duidelijk wordt.

Als het land van waaruit iemand werkt bepaalt of diegene überhaupt voor een functie in aanmerking komt, is locatie geen voetnoot. Het is onderdeel van de functievoorwaarden.

Woonland, werkland en nationaliteit 

Bij internationale recruitment worden begrippen als nationaliteit, woonplaats, fiscale residentie en werkland gemakkelijk door elkaar gebruikt. Juridisch zijn het verschillende aanknopingspunten.

Voor de sociale zekerheid geldt binnen de EU als uitgangspunt dat een werknemer verzekerd is in het land waar hij of zij daadwerkelijk werkt. Voor detachering en voor werknemers die structureel in twee of meer lidstaten werken bestaan afzonderlijke regels.

Ook het woonland kan daarbij relevant worden, bijvoorbeeld wanneer een substantieel deel van de werkzaamheden daar wordt verricht.

Het simpele label remote vertelt daarom niet onder welk socialezekerheidsstelsel iemand valt.

En dat raakt niet alleen de werknemer. Wanneer een werknemer onder het stelsel van een andere lidstaat valt, kan ook de werkgever daar verplichtingen krijgen rond registratie en premieafdracht.

Welk arbeidsrecht geldt? 

Sociale zekerheid en toepasselijk arbeidsrecht zijn afzonderlijke juridische vraagstukken.

Voor individuele arbeidsovereenkomsten binnen de EU is artikel 8 van de Rome I-Verordening van groot belang. Werkgever en werknemer kunnen in beginsel kiezen welk recht op hun overeenkomst van toepassing is.

Die keuze is echter niet onbeperkt.

Een rechtskeuze mag de werknemer niet de bescherming ontnemen van dwingende bepalingen van het recht dat zonder die rechtskeuze toepasselijk zou zijn geweest.

Zonder rechtskeuze wordt in beginsel gekeken naar het land waar of van waaruit de werknemer gewoonlijk werkt. Tijdelijk vanuit een ander land werken verandert die gewone werkplaats niet automatisch.

Een arbeidsovereenkomst waarin simpelweg staat “Dutch law applies” lost een grensoverschrijdende remote situatie dus niet noodzakelijk volledig op.

En als de werknemer verhuist? 

Hier wordt het juridisch bijzonder interessant.

Stel dat iemand al volledig remote werkt en tijdens het dienstverband naar een andere EU-lidstaat verhuist.

Operationeel kan er vrijwel niets veranderen. De werknemer doet hetzelfde werk, voor dezelfde werkgever, met dezelfde laptop.

Juridisch kan er wél veel veranderen.

Een verhuizing kan gevolgen hebben voor de sociale zekerheid, werkgeversverplichtingen en mogelijk voor de dwingende regels die de werknemer beschermen.

Daaruit volgt niet dat een werkgever iedere internationale verhuizing moet accepteren. Een onderneming kan legitieme redenen hebben om remote werken te beperken tot landen waarin zij werknemers juridisch en administratief kan faciliteren.

Maar andersom betekent een verhuizing ook niet automatisch dat een bestaande arbeidsovereenkomst zonder meer kan worden beëindigd. Dat hangt af van de overeenkomst, de afgesproken arbeidsplaats, het toepasselijke recht en de omstandigheden van het concrete geval.

Juist bij volledig remote functies is het daarom verstandig om vóór het sluiten van de arbeidsovereenkomst duidelijk te maken vanuit welke landen structureel mag worden gewerkt en wat er gebeurt wanneer een werknemer later wil verhuizen.

De digital nomad neemt zijn rechtspositie mee 

Remote werken heeft ook de term digital nomad populair gemaakt: iemand die digitaal werkt en zich geografisch relatief vrij kan bewegen.

Maar digital nomad is geen uniforme juridische status.

Een werknemer, zelfstandige of ondernemer neemt bij het oversteken van een grens niet simpelweg één onveranderlijke juridische positie mee in zijn laptop. Belasting, sociale zekerheid, verblijfsrecht en arbeidsrecht kennen ieder hun eigen regels en aanknopingspunten.

De vrijheid om ergens technisch te kunnen werken is daarom iets anders dan de juridische mogelijkheid om daar structureel te mogen werken.

Ook sociale bescherming kan veranderen 

Een verandering van werkland kan bovendien gevolgen hebben die pas veel later zichtbaar worden.

De Europese regels coördineren nationale socialezekerheidsstelsels, maar creëren geen uniform Europees stelsel. Welk nationaal stelsel van toepassing is, kan gevolgen hebben voor onder meer ziekte, werkloosheid en ouderdom.

Voor Nederlanders kan bijvoorbeeld de AOW relevant zijn. Volgens de Sociale Verzekeringsbank is iemand die buiten Nederland gaat wonen of werken meestal niet meer automatisch verzekerd voor AOW en Anw, hoewel belangrijke uitzonderingen bestaan. Voor ieder jaar waarin iemand daadwerkelijk niet voor de AOW verzekerd is, wordt in beginsel 2% minder AOW opgebouwd.

De keuze om structureel vanuit een ander land te werken kan dus tientallen jaren later nog financiële gevolgen hebben.

Misschien moet recruitment bij de grens beginnen 

Internationale remote recruitment begint nu vaak met de functie en eindigt pas later met de juridische vraag of de kandidaat vanuit zijn of haar land daadwerkelijk kan worden aangenomen.

Misschien moet die volgorde worden omgedraaid.

Een werkelijk duidelijke remote vacature zou al vóór de sollicitatie antwoord moeten geven op vier vragen:

Remote from where?

Employed where?

Resident where?

Taxed and socially insured where?

Dat voorkomt niet alleen teleurstelling bij kandidaten. Het bespaart recruiters en werkgevers tijd die anders wordt besteed aan kandidaten die juridisch of administratief uiteindelijk niet binnen de gekozen arbeidsstructuur passen.

“100% remote” is geen juridische categorie.

Het beschrijft vooral waar een werknemer níét hoeft te werken: op kantoor.

De interessante vraag voor de arbeidsmarkt van de toekomst is waar die werknemer juridisch wél kan werken.

Want als werk technisch niet langer aan een plaats gebonden is, wordt de paradox steeds zichtbaarder:

hoe locatie-onafhankelijk kan een arbeidsrelatie werkelijk zijn zolang het recht aan landsgrenzen gebonden blijft?

Joan Désirée Mulder, LL.M.

JAS – Aware Solutions

100% remote: hoe vrij ben je werkelijk?

100% remote: hoe locatie-onafhankelijk is een arbeidsrelatie juridisch? De juridische grenzen van werken zonder vaste werkplek 

Remote werken heeft de traditionele relatie tussen werk en plaats veranderd. Voor steeds meer werkzaamheden is een vaste kantoorlocatie technisch niet langer noodzakelijk. Een laptop, internetverbinding en toegang tot de systemen van de werkgever kunnen voldoende zijn om werkzaamheden vrijwel overal uit te voeren.

De opkomst van remote werken heeft ook de term digital nomad populair gemaakt: professionals die hun werkzaamheden digitaal en in beginsel vanaf verschillende locaties kunnen uitvoeren. Het begrip suggereert een grote geografische vrijheid, maar heeft op zichzelf nauwelijks juridische betekenis. Een werknemer, zelfstandige of ondernemer neemt zijn fiscale, sociaalrechtelijke en arbeidsrechtelijke positie niet simpelweg in zijn laptop mee. Zodra werkzaamheden landsgrenzen overschrijden, wordt relevant welk nationaal recht op welk onderdeel van toepassing is.

Dit artikel beperkt zich voornamelijk tot werknemers die binnen de Europese Unie grensoverschrijdend remote werken. Juist bij hen ontstaat een interessante tegenstelling: het werk kan technisch locatie-onafhankelijk zijn, terwijl de arbeidsrelatie juridisch sterk aan een land verbonden blijft.

Woonplaats, residentie en werkland zijn niet hetzelfde 

Bij internationaal remote werken worden woonplaats, residentie en werkplaats gemakkelijk als synoniemen gebruikt. Juridisch moeten zij echter van elkaar worden onderscheiden.

Voor de sociale zekerheid geldt binnen de EU als hoofdregel dat een werknemer onder de wetgeving valt van het land waar hij of zij daadwerkelijk werkt. Waar de werknemer woont of waar de werkgever is gevestigd, is daarvoor niet zonder meer bepalend.

Op die hoofdregel bestaan uitzonderingen. Een werknemer die tijdelijk naar een andere lidstaat wordt gedetacheerd, kan onder voorwaarden verzekerd blijven in het oorspronkelijke land. Ook voor werknemers die structureel in twee of meer landen werken gelden afzonderlijke regels. Daarbij kan bijvoorbeeld van belang zijn welk deel van de werkzaamheden in het woonland wordt verricht.

Het etiket "remote" geeft daarom nog geen antwoord op de vraag waar iemand sociaal verzekerd is.

Ook voor werkgevers is dat relevant. Wanneer een werknemer onder het socialezekerheidsstelsel van een andere lidstaat valt, kan de werkgever daar verplichtingen krijgen rond registratie en premieafdracht. Een buitenlandse werknemer betekent dus niet dat sociale verplichtingen verdwijnen. Zij kunnen juist naar een andere jurisdictie verschuiven.

Welk arbeidsrecht is van toepassing? 

Sociale zekerheid en toepasselijk arbeidsrecht zijn verschillende juridische vraagstukken.

Voor individuele arbeidsovereenkomsten binnen de EU is de Rome I-Verordening van groot belang. Artikel 8 bepaalt dat werkgever en werknemer in beginsel een rechtskeuze kunnen maken. Zij kunnen bijvoorbeeld overeenkomen dat Nederlands recht de arbeidsovereenkomst beheerst.

Die keuzevrijheid is echter niet onbeperkt. Een rechtskeuze mag een werknemer niet de bescherming ontnemen van bepalingen waarvan op grond van het recht dat zonder die keuze toepasselijk zou zijn geweest niet bij overeenkomst kan worden afgeweken.

Zonder rechtskeuze wordt in beginsel gekeken naar het land waar of van waaruit de werknemer gewoonlijk zijn werkzaamheden verricht. Tijdelijk werken in een ander land verandert die gewone werkplaats niet automatisch. Wanneer het toepasselijke recht langs deze weg niet kan worden vastgesteld, kijkt de verordening naar het land waar zich de vestiging bevindt waardoor de werknemer in dienst is genomen. Uit de omstandigheden kan bovendien blijken dat de overeenkomst nauwer verbonden is met een ander land.

Een contractuele bepaling dat Nederlands recht van toepassing is, betekent bij grensoverschrijdend remote werken dus niet automatisch dat uitsluitend Nederlandse regels relevant zijn.

Sociale bescherming kan veranderen 

De keuze om structureel vanuit een ander land te gaan werken kan daarnaast gevolgen hebben voor de sociale bescherming van een werknemer.

De Europese regels coördineren de nationale socialezekerheidsstelsels, maar maken daarvan geen uniform Europees stelsel. Het uitgangspunt is juist dat op een persoon op enig moment het socialezekerheidsstelsel van één land van toepassing is.

Dat stelsel kan relevant zijn voor onder meer ziekte, werkloosheid en ouderdom. Voor Nederlandse werknemers verdient daarbij ook de AOW aandacht. Volgens de Sociale Verzekeringsbank is iemand die buiten Nederland gaat wonen of werken meestal niet meer verzekerd voor de AOW en Anw, hoewel daarop uitzonderingen bestaan. Voor ieder jaar waarin iemand daadwerkelijk niet voor de AOW verzekerd is, wordt in beginsel 2% minder AOW opgebouwd. Onder voorwaarden is vrijwillige verzekering mogelijk.

Een beslissing over het land van waaruit structureel wordt gewerkt, kan daardoor veel verder reiken dan de keuze van een werkplek.

Wat gebeurt er wanneer een remote werknemer verhuist? 

Een bijzonder vraagstuk ontstaat wanneer een werknemer tijdens de arbeidsovereenkomst naar een ander land wil verhuizen.

Vanuit operationeel perspectief hoeft er niets te veranderen. Dezelfde werknemer kan met dezelfde laptop dezelfde werkzaamheden voor dezelfde werkgever blijven uitvoeren. Juridisch kan de verhuizing echter gevolgen hebben voor het toepasselijke socialezekerheidsstelsel, verplichtingen van de werkgever en mogelijk voor de dwingende regels die op de arbeidsrelatie van toepassing zijn.

Daaruit volgt niet dat een werkgever iedere internationale verhuizing van een remote werknemer moet toestaan. Werkgevers kunnen legitieme redenen hebben om werken te beperken tot landen waarin zij de arbeidsrelatie juridisch en administratief kunnen faciliteren.

Andersom betekent een verhuizing evenmin automatisch dat een bestaande arbeidsovereenkomst zonder meer kan worden beëindigd. Of beëindiging mogelijk is, hangt onder meer af van de arbeidsovereenkomst, afspraken over de arbeidsplaats, het toepasselijke recht en de ontslagregels die in het concrete geval gelden.

Wanneer Nederlands ontslagrecht van toepassing is, kan een werkgever een werknemer in beginsel niet zonder redelijke grond ontslaan. Afhankelijk van de ontslaggrond en omstandigheden kunnen bovendien herplaatsingsverplichtingen en toetsing door UWV of kantonrechter aan de orde zijn.

Juist bij volledig remote functies verdient het daarom aanbeveling om vooraf duidelijk vast te leggen vanuit welke landen mag worden gewerkt en wat de gevolgen zijn wanneer de werknemer later van woon- of werkland wil veranderen.

100% remote is geen juridische categorie 

Daarmee wordt ook de beperking van de term "100% remote" zichtbaar.

De term beschrijft vooral waar iemand niet hoeft te werken: op kantoor. Hij vertelt veel minder over de landen van waaruit iemand wél mag werken.

Een werknemer die volledig vanuit één vooraf overeengekomen land thuiswerkt, een werknemer die incidenteel enkele weken vanuit het buitenland werkt, een werknemer die structureel vanuit meerdere lidstaten werkt en een digital nomad die regelmatig van land wisselt, kunnen allemaal remote werken. Juridisch zijn dit echter verschillende situaties.

Voor werkgevers én werknemers is daarom meer duidelijkheid nodig dan alleen het label "100% remote". Relevante vragen zijn onder meer vanuit welke landen structureel mag worden gewerkt, welk socialezekerheidsstelsel van toepassing is, welk recht de arbeidsovereenkomst beheerst en welke gevolgen een latere internationale verhuizing kan hebben.

De ontwikkeling naar locatie-onafhankelijk werken is daarmee niet alleen een technologische of organisatorische verandering. Zij stelt ook het arbeidsrecht voor een fundamentele vraag:

Als werk niet langer aan een plaats gebonden is, hoeveel juridische betekenis moet die plaats dan nog hebben?