Global Internship Ecosystems

How cross border early career systems actually work, where they create value, where they quietly fail the people inside them, and what changes when trust becomes infrastructure.

Future Proof Intelligence. Research. No. VII. MMXXVI

Abstract

The internship is one of the most consequential transactions in a person's working life and one of the least examined as a system. Every year millions of young people cross a border, or a discipline, or simply the line between study and work, on the promise that a placement will give them a foothold. Beneath that promise sits a market with four layers that do not share an interest: a funding and policy layer of supranational programmes and national agencies, an institutional layer of universities carrying credit and duty of care, a host layer of organisations that absorb early career labour, and an intermediary layer that brokers access for a fee. The student passes through all four and is, structurally, the weakest party at every handoff.

This paper maps that system as it actually operates in 2026, grounded in the real landscape: Erasmus+, national internationalisation agencies such as Nuffic, the European regulatory turn of 2024, and the for profit placement economy that runs alongside the public one. It argues that the system reliably creates value at the placement itself and reliably captures it upstream, that it measures what is easy to count rather than what matters, and that trust inside it is carried by brand and paperwork rather than by any continuous, portable record of how actors have behaved. It then describes the missing layer: an intelligent substrate that holds identity, trust, and continuity of meaning across the handoffs, and regenerates the people inside the system rather than extracting from them. That substrate is the ground beneath the oldest operating vertical in this ecosystem.

1. The transaction nobody examines as a system

1.1 Why this is the right place to start

A research library that begins with the future of governance and the architecture of intelligence has an obligation, somewhere, to put its hands on something real and load bearing. The internship is that thing. It is not abstract. It is a nineteen year old in a city whose language she half speaks, on a train to an office that may or may not have prepared a desk for her. It is a small organisation in another country deciding whether the person arriving on Monday is going to be useful or a cost. It is a university administrator reconciling a placement against a credit requirement, a national agency closing out a funding line, a programme officer counting mobilities against a target.

Each of those people is acting rationally inside their own layer. None of them is responsible for the system that connects them. That is the first and most important observation in this paper, and almost everything else follows from it. The cross border internship is not designed. It is the emergent product of four layers optimising separately, and the person it is nominally for is the only actor present at every layer and the only one with no structural power in any of them.

We treat this not as a complaint but as an engineering description. Systems that no one designs still have a shape. They have load paths, failure modes, places where value concentrates and places where it leaks. The purpose of this paper is to draw that shape precisely enough that it can be reasoned about, and then to describe the layer the shape has been missing.

1.2 The scale, stated honestly

It is tempting to open with a large number. The honest position is more useful. The number of internationally mobile students, those who cross a border for tertiary study, rose from roughly 2.5 million at the start of the 2000s to approximately 6.9 million by the early 2020s, an increase of around 176 percent in two decades, with widely cited projections placing the figure above 10 million by 2030. These figures are aggregated principally by the UNESCO Institute for Statistics, which sits inside the United Nations system and functions as the primary international register of this flow.

That is degree mobility. The internship and traineeship layer, the work placement specifically, is measured far less well. There is no equivalent of the UNESCO register that counts, with confidence, how many young people cross a border for a placement each year, how many were paid, how many converted into anything, how many were structured at all. The structural truth is more telling than any invented precision: the placement layer, the part of the system where the lived value actually happens, is the part that is least counted. A system measures what it is built to optimise. The internship is not, at the system level, built to be optimised for the intern. The data gap is not an oversight. It is a fingerprint.

1.3 What a placement is for, and to whom

A placement does at least four different jobs depending on who is looking at it. To the student it is a foothold: experience, a reference, a signal, sometimes a route to a first contract, sometimes simply proof to themselves that they can do the work. To the university it is a curricular component, a credit, an employability statistic, a duty of care, a reputational surface. To the host it is labour, often early career and inexpensive, sometimes genuinely developmental, sometimes a way to test a hire without the cost of one. To the funding and policy layer it is a unit of mobility, an absorption of budget, a line in an evaluation, a contribution to a strategic objective such as cohesion or skills.

These four definitions are not contradictory in theory and are routinely contradictory in practice. The same eight weeks can be, simultaneously, a high quality developmental experience and an unpaid substitution for a junior salary; a successful mobility statistic and a placement that taught the person nothing; a glowing host testimonial and a relationship the student would not repeat. The system has no mechanism for resolving these into a single truth, because the system has no actor whose job is the single truth. This is the gap the rest of the paper is about.

1.4 A note on method

This paper is written as a structural analysis rather than a survey, and the choice is deliberate enough to state. There is a large descriptive literature on internships, much of it consisting of programme evaluations, satisfaction surveys, and policy reviews. That literature is valuable and this paper draws on it, but it shares a limitation: it tends to study the system from inside one of its layers, using that layer's instruments and that layer's definition of success. A programme evaluation asks whether the programme met its objectives. A university study asks whether placements improved its employability metric. A provider's own data asks whether its customers were satisfied at the point of sale. Each is honest within its frame, and none of them can see the system whole, because seeing it whole requires standing outside every layer at once, which is precisely the position no actor in the system occupies.

The method here is to take that outside position deliberately. We treat the cross border internship the way one would treat any system that no one designed: identify the components, trace the flows between them, locate where value is created and where it accumulates, find the load paths and the places with no load path at all, and ask what the persistent failures have in common. This is closer to the analysis of an infrastructure or a market than to the evaluation of a programme. It is the only vantage from which the central claim of the paper, that the failures are not separate problems but one missing layer seen from several angles, becomes visible rather than merely asserted. Where the paper makes an empirical claim it is grounded in the public record cited at the end. Where it reasons about structure it says so, and asks to be judged on whether the structure it describes is the structure that is actually there.

2. The four layers, drawn precisely

2.1 The funding and policy layer

At the top sits the layer that pays for and frames the flow without ever meeting the student. In the European context its most visible instrument is Erasmus+, the European Union programme for education, training, youth and sport, running in its current period from 2021 to 2027 with an overall indicative envelope of more than 26 billion euro. Roughly seventy percent of that budget supports mobility, with the remainder directed at cooperation and policy. Within mobility, Erasmus+ explicitly supports traineeships, that is work placements and internships, abroad for enrolled higher education students at every cycle and for recent graduates, with long term placements from two to twelve months, short term placements of five to thirty days, and blended options combining a short physical stay with a virtual period.

The important thing about this layer is not its size. It is its distance from the placement. The programme sets eligibility, duration bands, grant rates, and the documents that must exist for a placement to count: a learning agreement, a record of mobility, an evaluation. It does not, and structurally cannot, observe whether the placement was any good. It observes whether it was compliant. This is not a criticism of the programme, which has done more than any other instrument to make cross border early career experience normal rather than elite. It is a description of where this layer's competence ends. It can fund a placement and it can certify that it occurred. It cannot hold the relationship.

2.2 The national agency layer

Beneath the supranational programme sits a layer that almost no student can name and almost every placement passes through: the national internationalisation agency. The Netherlands is the clean example. Nuffic, the Dutch organisation for internationalisation in education, manages Erasmus+ for higher education at national level, including promotion, advice, project selection, monitoring and evaluation; it produces the cross border mobility research and data that informs Dutch policymakers; it runs the national front door for incoming students; and it studies questions such as the sustainability of internationalisation itself. Comparable bodies exist across the participating countries.

This layer is the connective tissue between an instrument designed in Brussels and an institution operating in a particular legal and linguistic context. It is genuinely valuable and genuinely necessary. It is also, like the layer above it, structurally upstream of the lived placement. Its instruments are programmes, datasets, and policy advice. Its unit of work is the cohort and the scheme, not the person and the relationship. When this layer is excellent, the pipes are clean and the data is honest. Even when it is excellent, it does not, and cannot, know whether the host treated the intern well, because that is not the kind of fact this layer is built to hold.

It is worth dwelling on what this layer is unusually good at, because the point of the paper is not that any layer is failing at its own job. The national agency layer is, at its best, the part of the system that produces honest knowledge about the system. The mobility research that bodies of this kind publish, on who goes where, on the balance of incoming and outgoing flows, on the sustainability of the whole enterprise, is among the few places where the system looks at itself with rigour rather than with marketing. That work is real and this paper relies on it. The structural observation is narrower and sharper for being narrow: the layer that knows the most about the system in aggregate knows almost nothing about any individual placement, and the layer closest to the individual placement, the host, knows almost nothing about the system. Knowledge in this system is distributed inversely to where it would do the most good. The agency sees the river and not the swimmer. The host sees the swimmer and not the river. No layer, and no instrument any layer possesses, sees both, and the student is the only party who is, by necessity, in both the river and the water.

2.3 The institutional layer

The university is the only layer above the placement that has a formal duty toward the individual student. It carries the credit the placement may satisfy, the academic recognition, often a legal and pastoral duty of care, and a reputational stake in outcomes. It is also, in most systems, the layer with the least capacity relative to its responsibility. A placement office may carry hundreds or thousands of students across dozens of countries and disciplines, with a handful of staff, a set of templates, and a relationship with each host that is frequently no deeper than an email thread and a signed agreement.

This produces a specific and well documented seam. The European Quality Framework for Traineeships, established by the Council Recommendation of 10 March 2014, sets out twenty one principles for traineeship quality, but its scope is deliberately bounded. It covers open market traineeships, those agreed directly between a trainee and a provider without a third party and generally undertaken after studies, and traineeships under active labour market policies. It explicitly excludes work placements that form part of formal education or vocational training curricula and traineeships regulated under national law as a mandatory route into a profession. In other words, the soft European quality instrument was deliberately drawn around the placements that universities do not run, leaving the curricular placement, the one with a university's name and duty of care attached, governed by national and institutional practice of widely varying quality. The layer with the clearest duty to the student operates, in quality terms, in the part of the map the common framework chose not to cover.

2.4 The host layer

At the bottom, where the work actually happens, sits the host: the organisation that takes the intern in. Hosts are not a monolith. A research institute hosting a doctoral trainee, a multinational running a structured graduate scheme, and a small studio taking on an unpaid pair of hands for a summer are all hosts, and the experience they produce is not the same experience. What they share is a structural position. The host is the layer that captures the most direct value from the placement, the work itself, and in a large part of the market the host pays the least for it. Evidence from the for profit placement economy, examined in the next section, shows providers that explicitly allow host organisations to take interns without paying any fee, with the cost borne entirely by the student.

The host is also the layer with the least accountability to anyone. The funding layer cannot see the host's behaviour. The agency layer does not measure it. The university has a duty of care but rarely the capacity to enforce it across borders. The intermediary, where one exists, is paid by the student and depends on host supply, which structurally discourages it from policing host conduct. The host, in short, is the layer that produces the experience and answers to almost no one for its quality. This is not a claim that hosts behave badly. Many are generous and developmental. It is a claim about the absence of a load path. Good host behaviour in this system is a matter of individual conscience, not of structural consequence.

This deserves to be made exact, because the asymmetry of consequence is the host layer's defining feature. Consider the two hosts a student cannot tell apart in advance. The first prepares for the person arriving: a defined piece of real work, someone responsible for them, a sense by the end of the placement that the student is more capable than they were. The second treats the placement as free or cheap hands: undefined tasks, no one accountable for the experience, a reference letter at the end that says nothing because there is nothing to say. In a system with a load path, the first host would accumulate something durable from its conduct and the second would pay a cost for its conduct, and over time the difference would compound. In the system as it exists, the first host's generosity evaporates with the student who experienced it, and the second host's extraction is invisible to the next student who arrives. The good host is not rewarded by the structure and the bad host is not penalised by it. Both are simply unobserved. A system that cannot distinguish its best participants from its worst is not neutral toward quality. It is, in effect, hostile to it, because it removes the only mechanism by which quality could compound: memory that travels and consequence that follows.

There is a second, quieter point about the host layer that the rest of the paper depends on. Hosts are not only the layer with the least accountability; they are also the layer with the most to gain from a system that remembered. A genuinely developmental host currently has no way to convert its own decency into anything that lowers its cost of attracting capable people, because its decency is private knowledge held by individuals who move on. The host that invests in its interns is, under the current structure, subsidising a public good it cannot capture. This is worth holding onto, because it means the missing layer described later in this paper is not a constraint imposed on hosts against their interest. For the good host it is the first mechanism that would let its conduct work in its own favour. The structure does not only fail the student. It fails the host that was trying to do the right thing, by making that effort indistinguishable, to the world, from the effort never made.

3. The intermediary economy

3.1 A market built on an asymmetry

Running alongside the public programme layer is a private one: a market of for profit placement providers that match students to host organisations for a fee. Its structure is consistent enough to describe as a model. The student, or the student's family, pays a programme fee, often with a deposit to confirm and a transaction surcharge, frequently bundling accommodation, visa support, and a placement guarantee. The host, in a significant part of this market, pays nothing. The vast majority of placements brokered this way are unpaid to the student.

Look at that flow as an engineer would. The party that pays is the party with the least bargaining power and the least information: a young person, often abroad for the first time, who cannot easily verify the quality of what they are buying until they are inside it and the money is spent. The party that benefits from the labour pays nothing and carries no fee based obligation to the intermediary. The intermediary's revenue is a function of student volume and host supply, not of placement quality, because quality is not what the paying party can observe at the point of sale. This is not an accusation against any particular provider. It is the description of an incentive structure. A market in which the payer cannot assess quality and the beneficiary bears no cost will, on average, optimise for throughput and reassurance rather than for the durable value of the experience. That is what the structure rewards, so that is, on average, what the structure produces.

3.2 What the intermediary actually sells

It is worth being precise about what is being purchased, because it is not, mostly, the placement. Direct application to a host is free. What the intermediary sells is the removal of three frictions: the friction of finding a host willing to take you, the friction of trusting a host you cannot verify from another country, and the friction of the logistics, visa, housing, and the sense that someone has your back if it goes wrong. The product is, in large part, the reduction of uncertainty for a frightened buyer. That is a real service. The problem is not that it is sold. The problem is that it is sold by a party whose incentives are not aligned with the buyer's actual interest, and that the uncertainty it reduces is uncertainty the system could have removed structurally and chose not to.

This is the single most important reframing in this paper. The intermediary economy is not a villain. It is a price the system charges for a missing layer. Every euro a family pays for reassurance is a measure of trust that the system failed to provide for free. The size of the intermediary market is, read correctly, a direct measurement of the absence of an honest, portable, verifiable trust layer in cross border early career. Where trust is infrastructure, the reassurance premium collapses, because reassurance is no longer a scarce private good. Where trust is absent, it is sold, and it is sold most expensively to the people least able to judge whether they are getting it.

It is worth pausing on the equity consequence of that last sentence, because it is not a rhetorical flourish but one of the best documented facts in this field. The research on unpaid and self financed early career experience is consistent and uncomfortable. Surveys across European countries have found that a large share of young people, in some countries well over seventy percent, have undertaken at least one traineeship, and that a substantial proportion of those are unpaid or insufficiently paid. The same body of work documents what unpaid placement does to who gets ahead. Desirable unpaid positions cluster among those who can afford to work for free, typically graduates from more privileged backgrounds or more prestigious institutions, while less desirable unpaid positions fall to those with no better option. The evidence further indicates that unpaid placements do not reliably improve employment prospects and are in several studies associated with lower subsequent earnings than no placement at all. Read together with the intermediary model, the picture sharpens into something stark: a market in which the reassurance and access that the public system fails to provide is sold privately, priced beyond the reach of exactly the people the public system was meant to serve, and delivering, for the unpaid placement, a return that the evidence says is at best uncertain and at worst negative. The intermediary economy does not merely charge for a missing layer. It charges for it regressively, and the missing layer's absence is therefore not a neutral inefficiency. It is a mechanism that converts an absence of public trust infrastructure into a private advantage for those who could already afford the experience and a private cost for those who could not.

3.3 The recognition and continuity problem

There is a quieter failure beneath the visible one. A placement, even a good one, tends to evaporate. The student finishes, receives at best a certificate and a reference letter, and carries forward a claim, "I interned at X for three months," that the next actor in their life cannot verify without starting again from zero. The host's view of the student does not travel with the student. The student's view of the host does not travel to the next student. The university's record, where it exists, is locked inside the university. The funding layer's record proves the mobility happened, not what it was.

Every handoff in this system re verifies from scratch or does not verify at all. The reference letter is the system's admission of its own statelessness: a single, unverifiable, easily inflated artefact standing in for a continuous record of how a person actually worked and how an organisation actually treated them. A system that cannot remember cannot improve, and cannot protect. The intern who is mistreated this summer has no way to warn the intern who arrives next summer. The host that is consistently excellent has no way to accumulate that excellence into anything that lowers its cost of attracting good people. The absence of memory is not a neutral gap. It is the precondition for the extraction the rest of this paper describes.

3.4 Recognition is the same problem wearing a formal name

It is worth connecting this to a question the institutional layer has spent decades on without resolving, because it shows that the statelessness diagnosed here is not a novel observation so much as an old one the system has been unable to fix from inside. The recognition of learning across borders, the question of whether a placement in one country counts toward a qualification in another, has its own elaborate apparatus: credit conventions, learning agreements, transcripts, frameworks designed precisely to make experience portable across institutional boundaries. The apparatus exists because the problem is real and well understood. A great deal of careful work has gone into making the formal envelope of a placement travel.

Notice what that apparatus recognises and what it cannot. It can recognise that a placement of a certain duration, against a certain learning agreement, occurred and was signed off. It carries the form forward across the border. It cannot carry forward what actually happened inside the form: whether the learning was real, whether the host honoured the agreement it signed, whether the person came back more capable or merely returned with a stamped document. The recognition apparatus is, in effect, a sophisticated mechanism for making the shell portable while the substance stays trapped where it happened. This is not a criticism of credit recognition, which does important work at the level it operates. It is the observation that the field's most developed instrument for portability is portability of the certified form, and that the thing the person on the train needs carried forward, the truth of how she was actually treated and what she actually became, is precisely the thing no recognition framework was built to hold. The recognition problem and the trust problem are the same problem. The system has spent its effort on the half of it that can be put on a transcript, because that is the half its instruments can reach, and left the half that matters most to the individual exactly where it has always been: nowhere, evaporating, unrecorded, the moment the placement ends.

4. Where the value is made and where it is captured

4.1 Value is made at the placement

The research on this point is unusually clear and worth stating without hedging. Work integrated learning, the family of practices that places study alongside genuine work, produces real and measurable gains in employability, job readiness, and access to graduate level employment. Peer reviewed studies find that the effect varies with the number of hours spent on placement and with the type of work integrated learning, and that placements crossing a border or a culture are associated with the adaptive skills employers most consistently say they lack in new graduates. The placement, done properly, works. This is not in serious dispute.

The same body of evidence carries a condition that the system routinely ignores. The gains accrue when the placement has genuine learning content and structure. The literature that demonstrates the value of work integrated learning also documents that low quality, unpaid, unstructured placements do not deliver these returns, and that unpaid placements are, in several studies, associated with no employment advantage and sometimes with lower subsequent earnings than no placement at all. The value is not in the existence of a placement. It is in the quality of one. The system, as drawn in sections two and three, measures existence and is structurally blind to quality. It therefore reliably produces the form of the thing that works while being unable to tell whether it has produced the substance.

4.2 Value is captured upstream

Trace where the value actually lands. The work the intern does is captured by the host, directly and immediately, often at little or no cash cost. The reputational and statistical value, a mobility completed, an employability metric improved, a programme objective advanced, is captured by the institution and the policy layer. The fee value is captured by the intermediary. The student captures what is left: the experience itself, which is real when the placement is good and close to nothing when it is not, and a reference letter whose worth the next actor cannot verify.

Set this against where the value is made. It is made at the placement, in the lived work and the relationship, by the student and the people immediately around them. The structural fact of this system is a separation between the site of value creation and the sites of value capture, with the direction of capture running upstream, away from the person doing the work. This is the precise technical meaning of extraction in this paper. Not malice. Not a conspiracy. A topology in which the layer that creates the value is the layer with the least power to retain it, and every layer above it captures a form of value it did not itself create at the placement.

The word extraction is used carefully here and should be defended, because it is easy to hear it as an accusation and it is not one. Extraction, in the sense this paper means, is a property of a topology and not of an intention. A river does not intend to carry silt downstream; the gradient does that, and the gradient is a fact about the landscape, not a moral failing of the water. The gradient in this system is the direction in which value can be retained, and it runs upstream because only the upstream layers have the durability to hold value over time. The host can hold the labour because the host persists. The institution can hold the statistic because the institution persists. The intermediary can hold the fee because the intermediary persists. The student cannot hold the experience as anything transferable because the student, in this system, does not persist as a remembered entity at all; the moment the placement ends, the system's memory of the student ends with it. Extraction here is simply what happens when value is created by the only party the system does not let persist. Correct the persistence and the gradient changes, not because anyone became more virtuous, but because the landscape was rebuilt. This is why the paper insists, repeatedly, that the problem is structural. A structural problem has the unusual and hopeful property that it can be solved structurally, without requiring any participant to be better than they are.

4.3 The measurement trap

Why does this persist when the harm is documented and the beneficiaries are, individually, often well intentioned? Because the system measures what is easy to count and governs by what it measures. Mobilities are countable. Budget absorption is countable. Enrolment and completion are countable. Whether a placement held a person, taught them something true, changed the host that took them in, or sent them home more capable and less afraid: none of that is countable with the instruments the system has, so none of it governs behaviour.

This is the measurement trap, and it is general to systems of this kind, but it is acute here because the uncounted variable is a human being's first serious encounter with working life. A national agency that is rewarded for mobilities will, rationally, produce mobilities. A university rewarded for an employability statistic will produce the statistic. An intermediary rewarded for placements will produce placements. Every actor optimises honestly for its metric, and the sum of honest local optimisation is a system that is excellent at producing the measurable shell of early career experience and indifferent, structurally, to whether the experience inside the shell was real. You do not fix this by exhorting the actors to care more. You fix it by changing what the system can see.

5. The regulatory turn, and what it does not reach

5.1 What is hardening now

The roots of this system are being set in regulation as this paper is written. On 20 March 2024 the European Commission proposed a two part package: a Traineeships Directive aimed at improving and enforcing working conditions for trainees and at combatting regular employment relationships disguised as traineeships, and a revised Council Recommendation reinforcing the 2014 Quality Framework on questions of fair pay, social protection, and inclusiveness. The Commission's own 2023 evaluation had found the 2014 Recommendation positive in impact but insufficient in reach. Once a directive of this kind is adopted by the co legislators, Member States have a defined period, on the order of two years, to transpose it into national law.

This matters beyond its content. It tells us which questions the system has decided are real enough to legislate. Disguised employment is being legislated. Pay and social protection are being legislated. Enforcement of working conditions is being legislated. These are the right questions and the package is a genuine advance. But notice the shape of the answer. The binding instrument addresses the abuse of the placement as a labour relationship. The soft instrument addresses the quality of the placement as an experience. Neither instrument addresses the statelessness identified in section three: the absence of a portable, verifiable, continuous record of how actors actually behave across handoffs and over time. The regulation is hardening around compliance and pay. It is not hardening around trust and memory. That seam is being left open precisely as the rest of the structure sets.

5.2 Why compliance is necessary and not sufficient

A compliance regime can require that a placement be paid, that hours be bounded, that a learning agreement exist, that a relationship not be a disguised job. It cannot, by its nature, make a placement good. It governs the floor, not the substance. A perfectly compliant placement can still be empty: legally paid, correctly documented, and developmentally worthless. Compliance removes the worst of the downside. It does nothing to produce the upside, and it generates its own characteristic failure mode, the placement engineered to pass the audit rather than to serve the person, which the measurement trap of section four will then record as a success.

This is the limit of the regulatory turn, and stating it is not an argument against regulation. The floor matters and raising it protects real people from real harm. The argument is narrower and more structural: a system whose roots harden around compliance alone will have, a decade from now, a well regulated version of the same topology. The extraction will be legal. The reassurance will still be sold to the frightened buyer, now with a compliance badge attached. The memory will still not exist. Compliance is a necessary root. It is not the root that changes the shape.

5.3 The window

There is a timing observation here that is the quiet spine of this whole library. The structure of a market is most malleable in the period when its rules are being written and least malleable afterward. We are inside that period for cross border early career. The instruments being drafted now will define what counts, what is enforced, and what is simply outside scope, for the working lives of the people currently in school. A trust and continuity layer that is built into the foundation while the foundation is wet becomes load bearing. The same layer proposed after the concrete has set is an aftermarket accessory, argued about and rarely adopted. The cost of the missing layer is not constant. It is rising, on a schedule, as the rest of the system sets around its absence.

6. What the system is actually optimising for

6.1 Three things called the same word

Before naming the missing layer it is worth resolving an ambiguity that runs through every layer described so far and quietly causes most of the dysfunction. The word the whole system uses for what it produces is the same word, internship, but the three things that word refers to are not the same thing, and the system's failure to distinguish them is not an oversight of language. It is the mechanism of the harm.

The first thing is a labour relationship: a young person doing work an organisation needs, for which the relevant question is whether they are paid fairly, treated lawfully, and not substituting for a job that should exist. This is the thing the 2024 regulatory package is built to govern, and it is right to govern it.

The second thing is a learning relationship: a structured period in which the point is not the work the organisation extracts but the capability the person develops, for which the relevant question is whether the placement had genuine content, a defined arc, and someone responsible for the learning rather than only for the output. This is the thing the work integrated learning literature shows produces real returns, and the thing the system is least equipped to verify.

The third thing is a trust relationship: a person entering, often across a border and a language, an organisation they cannot assess in advance, on the implicit promise that they will be held rather than used, for which the relevant question is whether anyone remembers how this organisation has actually behaved toward the people it took in before. This is the thing no instrument in the system governs at all.

Every placement is all three at once. The system has a maturing instrument for the first, a weak and optional instrument for the second, and no instrument for the third. When it says a placement was successful, it almost always means the first was lawful and the paperwork for the second exists. It has no way to even ask about the third, which is the one the person on the train actually cares about, because the third is the only one of the three that is about whether the system can be trusted with them.

6.2 Why measurement, not motive, is the binding constraint

It is tempting to locate the system's failure in the motives of its actors: the exploitative host, the metric chasing institution, the extractive broker. This paper has argued the opposite throughout and the point is worth consolidating. The binding constraint is not motive. It is measurement. A system governs by what it can see, and what it can see is determined by what it can measure, and what it can measure in cross border early career is occurrence and compliance, not quality and trust. Improve every actor's motives overnight and the system would still produce, on average, the same outcomes, because a well intentioned actor optimising against a metric that cannot see quality still cannot produce quality reliably; they can only produce it by private effort that the system neither rewards nor remembers.

This is why exhortation, ethics charters, and codes of good practice, all of which exist in this field and none of which have changed its shape, do not work. They address motive in a system whose constraint is measurement. The only intervention that changes the shape of a measurement bound system is one that changes what the system can see. Everything else is a moral appeal to actors who are, mostly, already as well intentioned as the structure allows them to be, and structurally prevented from acting on it in a way that compounds. The conclusion is not pessimistic. It is the opposite. It means the problem is not the people, and a problem that is not the people is a problem that can be engineered.

7. The missing layer

7.1 Naming it precisely

Everything to this point has been a structural diagnosis with no reference to any particular actor. The diagnosis converges on a single absence. The system has a funding layer, an agency layer, an institutional layer, a host layer, and an intermediary layer. It does not have an identity and trust layer: a continuous, portable, verifiable substrate beneath all five that remembers how actors have actually behaved, carries that memory honestly across every handoff, and is governed so that it cannot be captured by any one layer's interest.

Call this the identity layer over the orchestration layer. The orchestration layer is everything the system already has: the programmes, the agencies, the universities, the hosts, the brokers, the matching, the logistics, the documents. Orchestration is, increasingly, a solved and commoditised problem; software can match a student to a host and move the paperwork. The scarce and durable thing is not the orchestration. It is the layer beneath it that holds who an actor is, how they have behaved, and what their participation has actually meant, continuously and across boundaries, in a form the next actor can trust without starting from zero. The internship system's defining failures, extraction, the measurement trap, the reassurance premium, the statelessness, are all symptoms of the same missing layer. They are not separate problems. They are one problem seen from four angles.

7.2 What the layer does, structurally

Described only at the level of principle, because principle is what is solid here, the identity and trust layer changes four things at once.

It makes behaviour remembered. A host that treats people well accumulates that as a portable, verifiable property rather than as the private knowledge of whoever happened to be there. A student's actual work travels with the student rather than evaporating into an unverifiable letter. The next handoff inherits memory instead of re verifying from scratch or not at all.

It moves trust from brand to record. Today trust is carried by the reputation of an institution or the marketing of a provider, both of which are claims. A trust layer carries trust as an accumulated record of conduct, which is evidence. The reassurance the intermediary economy sells expensively to frightened buyers becomes a property of the system rather than a private good, and the premium charged for it collapses toward its true cost, which is near zero.

It changes what the system can see, and therefore what it optimises. A layer that holds the quality of the relationship, not merely its occurrence, makes the previously uncountable countable in an honest way, and the measurement trap loosens because the metric and the meaning are no longer divorced. This must be built with the trauma the system currently inflicts in full view; a memory layer that records people can also expose them, and the design obligation is to hold the person, not to surveil them.

It regenerates rather than extracts. A system with memory can route value back toward where it was created. The student who did real work, the host that was genuinely developmental, the institution that protected its people: each can accumulate standing that compounds in their favour, rather than watching the value they created flow upstream and disappear. The topology of extraction described in section four is not a law of nature. It is a consequence of statelessness. Remove the statelessness and the direction of value can be designed rather than merely suffered.

7.3 Why this is a substrate and not a product

A natural misreading of section 7.2 is to imagine a single application that does these things. That is the wrong mental model and worth correcting directly. A product sits on top of the orchestration layer and competes inside it; it is one more intermediary, and an intermediary cannot fix a problem that is structural to the intermediary position. The thing the diagnosis points at is underneath the orchestration layer, not on top of it. It is a substrate: the layer the other layers stand on, referenced rather than sold, assumed rather than announced, the way one assumes the existence of a registry or a standard rather than buying it as a feature.

This is the register in which Future Proof should be understood in this paper, and the only register in which it appears. The argument of this library is that the durable layer of the intelligence era is not orchestration, which is becoming abundant, but identity, trust, and continuity of meaning across systems, which remains scarce. Cross border early career is simply the place where that thesis is the most concrete and the most human. It is the oldest operating vertical beneath this ecosystem not because internships are the point but because they are where the missing layer's absence costs a real person something real, every summer, at scale. The work of building the identity layer beneath this system has not been waiting for a paper to propose it. The paper exists because the building came first, and a serious foundation documents the ground it already stands on.

8. Implications

This section is deliberately divided by reader, because the same structural fact lands differently depending on where you stand in the system. None of what follows is a pitch. It is the diagnosis applied.

8.1 For institutions

You hold the clearest duty of care in the system and, relative to that duty, the least capacity and the least memory. The curricular placement, the one with your name on it, sits in the part of the quality map the common European framework deliberately did not cover, which means its quality is governed by your own practice and almost nothing else. The implication is not to do more with the same instruments. It is that an institution's duty of care is, in practice, a memory problem: you cannot protect a student in a placement you cannot see, and you cannot see a placement you do not remember across cohorts. The institutions that will hold their people well in the next decade are the ones that treat continuity of trust about hosts and placements as core infrastructure rather than as an administrative residue, and that recognise this substrate is something to stand on rather than something to rebuild alone.

8.2 For investors

The intermediary economy is frequently read as the investable surface of this market, because it has a clear fee and a clear customer. This paper's structural reading is the opposite. The intermediary fee is a measurement of a missing layer, and value built on the persistence of an asymmetry is value exposed to the closing of that asymmetry. The durable position in this system is not a better broker. It is the trust substrate beneath all the brokers, the layer whose value compounds with use and with memory rather than with throughput, and which becomes more load bearing precisely as the regulatory roots harden around everything except it. The relevant question for capital is not which intermediary wins. It is which layer the system cannot, in ten years, function without, and who was inside that layer before it set.

8.3 For operators

If you run a host, a programme, an agency, or a placement function, the actionable reading is narrow and direct. Every actor in this system is optimising honestly for a metric that is not the thing that matters, and the sum of that honest local optimisation is the system's failure. You cannot fix this alone by caring more, and you should not try to; the design literature is clear that exhortation does not change a topology. What changes it is participating in a layer that makes the right thing the remembered thing, so that good conduct accumulates into standing rather than evaporating into a reference letter. The operators who will compound advantage are the ones who treat their own record of conduct as a portable asset and choose to operate on a substrate that carries it honestly.

8.4 For the people inside these systems

This is the reader the system was supposed to be for and structurally is not. The most important thing this paper can say to a person about to cross a border for a placement is that the fear they feel at the point of paying for reassurance is rational, because the system genuinely does not remember and genuinely cannot, today, tell them whether the host will hold them or use them. That is not their failure to research hard enough. It is a missing layer, and its absence has been priced and sold back to them. The change that matters to this reader is not advice. It is a system that remembers on their behalf, that carries what they actually did forward as something true and theirs, and that is built to hold them rather than to extract from them. A foundation that does that is not a service they purchase. It is ground under their feet that should simply be there, the way ground is.

8.5 For the system as a whole

The four readings above are deliberately separated, but they converge on a single observation that is worth stating plainly because it is the practical core of the paper. The interests of the well intentioned actor in every layer point the same way once the missing layer exists, and point in conflicting directions while it is absent. Today the institution protects itself by managing reputational risk, which is not the same as protecting the student. The good host invests in people and captures nothing, which makes the investment irrational at scale. The honest intermediary competes on reassurance against rivals who can simply claim more of it, because no one can verify the claim. The student pays for trust that does not compound into anything they keep. Each actor is doing the locally sensible thing and the sum is the dysfunction the paper has described.

A trust and continuity substrate does not require any of these actors to act against their interest. It realigns their interests by changing what persists. The institution that genuinely protects its people accumulates a verifiable record of having done so, which is worth more than reputational management because it is evidence rather than claim. The good host's investment finally compounds, because its conduct toward people becomes a portable asset rather than a private cost. The honest intermediary can finally win on something verifiable instead of something asserted. The student keeps what they earned. This is the test of whether a proposed layer is real infrastructure rather than another product competing inside the system: real infrastructure makes the right thing the rational thing for everyone at once, without anyone having to be persuaded to be better than they are. That is the standard against which the missing layer should be judged, and the standard the substrate beneath this vertical is built to meet.

Coda

There is a nineteen year old on a train this morning, in a country whose language she half speaks, going to an office that may or may not have prepared a desk for her. Every layer of the system she is inside has done its job. The programme funded her. The agency cleaned the pipes. The university signed the agreement. The intermediary, if there was one, was paid. By every metric the system can see, she is a success before she has arrived.

The system cannot see the only thing that matters about her morning, which is whether the place she is going to will hold her or use her, and whether anyone will remember the difference. That is the whole argument of this paper compressed into one person on one train. The internship is not a small subject. It is the place where the largest thesis in this library, that the scarce layer of the intelligence era is not orchestration but trust, identity, and continuity of meaning, stops being abstract and becomes a person whose first encounter with working life is being decided, right now, by a system that does not remember her.

Build the layer that remembers, and the train ride is the same train ride but the person on it is no longer alone in it. Do not build it, and the system will be, a decade from now, a more compliant version of exactly this: excellent at producing the measurable shell of an experience, structurally indifferent to whether anyone was inside it. The roots are setting now. What gets built underneath them, while the ground is still soft, is the only part of this that is still ours to decide.

References and Notes

The following sources are real and publicly verifiable. Where this paper states a structural truth rather than a precise figure, it does so deliberately, because the precise figure either does not exist at the necessary resolution or is not solid enough to cement.

1. UNESCO Institute for Statistics. Internationally mobile students, glossary and data browser. The primary international register of cross border tertiary student flows, part of the United Nations system. (uis.unesco.org)

1. International Organization for Migration. World Migration Report 2024, chapter on international students. Corroborates the scale and growth of internationally mobile students. (worldmigrationreport.iom.int)

1. European Commission. Erasmus+ Programme 2021 to 2027, Programme Guide and budget pages, including the overall indicative envelope of more than 26 billion euro and the mobility share of the budget. (erasmus-plus.ec.europa.eu)

1. European Commission. Traineeships abroad for students, Erasmus+ opportunities pages, on the duration bands and eligibility for traineeships abroad. (erasmus-plus.ec.europa.eu)

1. Nuffic. Organisation pages and cross border mobility research, on the role of the national internationalisation agency in managing Erasmus+ for higher education, producing mobility data, and supporting incoming students. (nuffic.nl)

1. Council of the European Union. Council Recommendation of 10 March 2014 on a Quality Framework for Traineeships, CELEX 32014H0327(01), including the twenty one principles and the explicit scope covering open market traineeships and active labour market policy traineeships. (eur-lex.europa.eu)

1. European Commission. Proposal package of 20 March 2024: the proposed Traineeships Directive on improving and enforcing working conditions and combatting disguised employment, and the proposal to revise the 2014 Quality Framework, COM(2024) 132 and COM(2024) 133, with the related 2023 evaluation finding. (eur-lex.europa.eu, employment-social-affairs.ec.europa.eu)

1. International Labour Organization. The regulation of internships: a comparative study; and Interns Revealed, a survey on internship quality in Europe. On prevalence, unpaid placement, weak learning content, and weak conversion. (ilo.org)

1. European Youth Forum. Report on national internship policies (High Quality or Unpaid and Unregulated). On the prevalence of unpaid and unregulated internships and the equity consequences. (youthforum.org)

1. Peer reviewed work integrated learning literature, including studies published in Higher Education Research and Development on the contribution of different types of work integrated learning to graduate employability. On the documented gains from quality placements and the dependence of those gains on genuine learning content and structure. (tandfonline.com)

Note on the for profit placement economy. The structural description of the intermediary model in section three (student paid programme fees and deposits, host organisations frequently paying no fee, the majority of brokered placements unpaid to the student) reflects the publicly stated terms and operating models of the for profit placement provider market as observable in 2026. This paper does not name individual providers, and treats the model structurally rather than as a claim about any specific company.

Note on figures. Student mobility counts are reported as orders of magnitude and direction (approximately 2.5 million rising to approximately 6.9 million over two decades, an increase of roughly 176 percent, with cited projections above 10 million by 2030) because that is the resolution at which these figures are solid. The placement and traineeship subset is, as argued in the paper, materially less well measured, and no precise global placement figure is asserted here for that reason.

Future Proof Intelligence. Research. No. VII. MMXXVI.

Future Proof Intelligence . Research . No. VII . MMXXVI