"If I get into an accident, what will happen? The company will not pay anything for me." A rider said that in Nicosia in September 2025, at the European Trade Union Confederation's Platforum gathering, describing work that paid €8–10 an hour before tax. 1
On paper, the European Union had already answered him. On 14 October 2024, the member states' own governments, voting in the Council, adopted Directive (EU) 2024/2831 on improving working conditions in platform work, closing nearly three years of contested negotiations. No government voted against; only Germany abstained, over a split inside its own coalition. 2 3 4 The directive's centrepiece is a rebuttable presumption of employment: where the facts show a platform directs and controls the work, employment is presumed, and the platform, not the worker, must prove self-employment. 5 6 Platforms must disclose their algorithmic management systems; significant automated decisions require human oversight and a right to contest; processing that infers a worker's emotional state or predicts trade-union activity is banned. 7 Tens of millions of people work through EU platforms. 8 9 These are protections the Union decided, together, that those workers deserve.
Now count the member states that have fully transposed the directive as of late September 2026: zero. One, Slovakia, has passed a transposing law, approved at third reading on 23 September and applying from the deadline itself. Six (Belgium, Croatia, Italy, Malta, Portugal and Spain) have older national rules that cover part of it, mostly a presumption for some platform workers. Eleven have a bill somewhere in the pipeline, among them France, which has already said it will miss the deadline. Nine, Germany among them, have published no bill at all. 10
The deadline has not passed, and capitals often legislate in the final weeks. But nothing about this is unusual. Between 2012 and 2023, member states missed 5,902 of 20,076 transposition deadlines — almost three in ten. 11 Platform work is not an exception. It is an unusually clear view of how the system normally works: the Union decides, and then, file by file, capitals un-decide by waiting.
[image: A clock formed by the EU's ring of twelve gold stars on EU blue, its hands pointing to 2 December, with a gold 0/27 scoreboard beneath]
Agreed in Brussels. Fully in force in zero capitals.
How the Union un-decides what it has decided
A directive, in the words of the treaties, is binding "as to the result to be achieved" but leaves "the choice of form and methods" to national authorities. 12 In practice: agree the end in Brussels, then hand the means to twenty-seven capitals, each working on its own calendar. What holds them together is a deadline, and an enforcement machine that only starts once the deadline has passed.
For this file, the machine is already being warmed up. The Commission has said no postponement is possible 13 and has "quietly warned that infringement proceedings will follow any government that misses the December deadline". 14 The forecast on record is blunt: "most of the EU-27 will arrive at the deadline without ready national implementation, leading to a likely phase of national-court-driven interpretation and Commission infringement proceedings through 2027". 15 Even before any case reaches the Court of Justice, the Commission's handling of late-transposition cases has averaged around a year and a half since 2018. 16 Only after a judgment can the Court impose a lump sum or daily penalty. 17
Meanwhile the directive does very little for the people it was written for. Under settled case law, an untransposed directive can be invoked against the state, but not against a private company (Marshall, 1986; Faccini Dori, 1994). 18 Platforms are private companies. On 3 December, a rider in Paris disputing his status with an app cannot rely on the directive's presumption against that app. National courts must read existing law in the directive's light where they can, but they cannot rewrite a statute that points the other way — and France's statute points the other way. In theory the rider can sue the French state for damages (Francovich, 1991): years of litigation to recover what a statute should have given him on the first day.
So notice what the machine cannot do: deliver the law on time. It can open a case, win a judgment and fine a treasury years later. The people who wait are the workers. A sanction that arrives after the harm is not a deterrent; it is an invoice — and invoices are budgeted.
What nullification by delay looks like
France is the pure case. Its existing framework presumes registered platform workers to be independent, and France "resisted a strong presumption during negotiations", so transposition means unwinding a system it built itself. 19 Labour Minister Jean-Pierre Farandou told AFP on 11 September that the directive cannot be transposed before 2 December and is unlikely to pass before the presidential term ends: "My goal is that the bill exists. It's already not bad. There is a text, we can launch it but I fear that we will be caught up by a parliamentary framework that will make it difficult to vote on it before February." 20 The parliamentary session ends in February and the presidential election falls in April 2027. Against that calendar, the cost of missing the deadline is infringement proceedings and perhaps, eventually, a fine. 21 Known, priced, accepted. Brahim Ben Ali of FO-INV called the delay "unacceptable": "Drivers and delivery drivers have been waiting for years for real rights over algorithms, their data, automated decisions and their status." 22 The election calendar wins, and the decision taken together waits.
Germany shows the quieter form of the same veto. The government confirmed in April 2026, in answers to parliamentary questions, that the labour ministry is drafting; by 29 September, no ministerial draft bill had been published. 23 On 25 September the Bundesrat called on the government to bring one forward: a political demand, not a law. 24 25 Underneath sits the real fight. Unions demand a direct-employment mandate banning subcontracting, "following the example of the meat industry in Germany"; food-delivery platforms lobby against it; and the coalition reportedly leans toward a narrower triggering test. 26 27 None of that quarrel will be settled before the deadline it is quarrelling about.
The long tail repeats the pattern in smaller type. Czechia's bill proposes an effective date of 1 January 2027, after the deadline. Portugal's parliament rejected, in June, the labour reform that carried its transposition. Italy has applied a rider presumption since May 2026, but its full decree is stuck in committee. Estonia's bill is scheduled for a first reading only on 14 October. Sweden is under a caretaker government. Slovenia and Ireland have published nothing. 28
Spain is the contrast. Its rider presumption has been law since August 2021. 29 On 5 August 2026 it opened a consultation to extend that presumption from food delivery to all platform work, aiming to finish by year's end. 30 Spain may miss 2 December too, but its riders are already covered, because Spain settled the underlying argument on its own five years ago. The six states with partial rules in force tell versions of the same story: where the politics are done, transposition is a technical act. Where it lags, the cause is less often incapacity than unfinished politics, and the directive method lets those politics run past the deadline at no immediate cost.
Flip the default
Every directive with a transposition deadline should come with a base law: a complete, ready-to-apply text implementing the directive, negotiated and adopted by the Council and the European Parliament together with the directive itself. Where the directive leaves a choice to member states, such as which facts trigger the presumption, the base law makes that choice, and the co-legislators vote on it in the open.
Member states keep the window they have always had. They can adopt the base law as it stands, adapt it to national structures, or go further: a German parliament that wants a direct-employment mandate can pass one; Spain can keep extending its own design. The mechanism makes legislating the attractive option: take the agreed text, fit it to national needs, keep the credit, keep the pen.
What comes off the menu is doing nothing. If the deadline passes without national measures, the base law applies automatically in that member state, as adopted alongside the directive. No infringement case, no fine, no years of waiting. The penalty for inaction stops being a fine paid by the treasury; it becomes the loss of the pen.
Three effects follow.
Delay stops being leverage. Today a French election calendar can hold workers' rights hostage. Under a default, missing the deadline does not postpone the law; it gives up the country's say over the wording.
Divergence narrows. "How each country defines the criteria that trigger the employment presumption will decide how many of those workers gain holiday pay, sick leave and minimum-wage guarantees, and how many stay outside them." 31 The negotiations left those criteria "largely to individual Member States", which is why the same working pattern "will be classified differently across member states once the directive lands". 32 33 A base law gives every state the same working floor from the first day. The directive already forbids going below it; parliaments remain free to go further. What disappears is the vacuum in between.
Defaults are easier to agree than mandates. The Commission's 2021 proposal set out EU-wide criteria for triggering the presumption. 34 They were negotiated away because they would have bound every member state. A default that any parliament can legislate around asks far less of a reluctant government. The criteria could have survived as the base law's content, applying wherever a capital failed to replace them. Every future negotiation would face the same, easier question: not "will you accept this everywhere?" but "will you accept this where your own parliament chooses not to act?"
Today, the default result of national inaction is a vacuum. It should be the law.
How it could work
The obvious question is why the EU does not simply use a regulation, which applies directly in every member state with no national transposition at all. For labour law, the treaties rule that out. Article 153(2)(b) TFEU allows the Union to set minimum requirements for working conditions only "by means of directives". 35 That is why the employment presumption arrived as a directive in the first place, and why the delay problem is built into EU social policy rather than being a quirk of this file. (The directive's data-protection provisions also rest on Article 16 TFEU, which does permit regulations. The GDPR already shows that a directly applicable text with room for national rules can work.)
That leaves two routes.
The first needs no treaty change. Each member state passes, once, a standing default-transposition act: any directive whose deadline passes without national measures takes effect in national law in the form of its adopted base law, until parliament replaces it. This is not exotic. Italy already processes EU directives through an annual European delegation law; a default act makes the same idea automatic. A group of willing states could adopt one now, and the Commission could make the case for it in every annual enforcement report. The weakness is obvious: the governments most comfortable with delay are the least likely to sign up.
The second closes that gap: a treaty amendment allowing a directive to carry a base law that applies directly in any member state from the deadline until national measures are notified. Treaty change needs unanimity and national ratification, so this is a long fight. But treaty reform has been on the table since the European Parliament proposed amendments in 2023, in the debate over enlargement, and this amendment has an unusual selling point: it takes no power away from any member state. It changes only what happens when they don't use the power they have.
Either way, a base law has to fit twenty-seven legal systems. It cannot name the labour inspectorate in Ljubljana or the labour court in Lyon, so it should do what every EU regulation already does: state the rights and duties, assign enforcement to whichever national bodies already enforce labour law, and set a default penalty scale. That is a drafting problem, and the EU solves it every time it adopts a regulation.
The objections, taken seriously
Subsidiarity. Parliaments lose no power. They keep every ability to legislate, adapt or go further; they lose only the ability to block the floor by failing to act. The alternative, protecting national discretion by leaving a vacuum, is what France is demonstrating now: a minister weighing a future fine against an election calendar while drivers wait.
Legitimacy. "Law drafted in Brussels" is the wrong description. The base law would be voted on by the same national ministers in the Council and the same elected MEPs who adopt the directive, and it would bind only where a national parliament declined to produce a law of its own. A worker in Paris or Berlin does not experience that silence as sovereignty.
"Just enforce harder." Faster infringement procedures and earlier fines would help at the margin. But a faster invoice is still an invoice. It reaches the treasury, not the worker, and it still arrives after the deadline it is meant to protect.
The platforms' objection is the most honest one: platforms would lose the veto by delay but keep every argument in court. The reading that "any presumption in favour of an employment relationship will in many cases be rebutted by the circumstances as a whole", and the wider flexibility debate, survive under any mechanism. 36 37 38 A base law does not end that fight. It means the fight is about one text rather than twenty-seven gaps.
A base law also fixes timing, not everything. Five years of Spain's Rider Law produced "organisational restructuring, algorithmic opacity, and performative compliance": enforcement that "has not transformed the labour process but has instead generated a new regulatory equilibrium". 39 In Germany, the Federal Labour Court held in January 2026 that app-based delivery areas do not count as an "establishment" for electing a works council. 40 A statute is where enforcement begins, not where it ends. The base law's job is to make the floor arrive on time; making it hold is the work that comes after.
The decision we already made
The unions set their yardstick in Nicosia more than a year before the deadline: a working presumption, regulation of algorithmic management, and collective bargaining "to prevent the further use of algorithms as a union busting tool". 41 Every national law should be measured against that list. As ETUC's Tea Jarc put it: "Millions of riders, drivers and carers have already waited too long to receive their most basic rights as workers, such as paid sick leave or wages which don't fall below the legal minimum." 42
Two years after adoption, the scorecard reads: no government against, zero full transpositions, a French minister forecasting next year at the earliest, and a Bundesrat resolution asking Berlin to start. 43 44 45 46 Brussels is already drafting the sequel: Commissioner Roxana Mînzatu has been asked to look ahead to an initiative on algorithmic management and possible legislation on AI in the workplace. 47 Every such law will hit the same wall unless the mechanism changes.
Europe does not lack the capacity to decide. It lacks a way to keep what it decides. It should stop waiting for twenty-seven capitals to finish what they voted through, and start writing the version that applies itself.