A deadline that is closer than it looks
Directive (EU) 2024/2831 on improving working conditions in platform work must be transposed into national law by 2 December 2026. That is less than five months away, and most member states – Sweden included – are still working on their implementing legislation.
The directive is usually reported as a gig-economy story about food couriers and ride-hailing drivers. That framing is misleading. Its algorithmic management rules apply to digital labour platforms of every kind, and its logic – that automated decisions about people at work need transparency and human oversight – is spreading well beyond platforms.
The two halves of the directive
1. A presumption of employment
Where facts indicate control and direction, national law must presume that the relationship between a person performing platform work and the digital labour platform is an employment relationship. The presumption is rebuttable, but the burden falls on the platform, and the platform must prove that no employment relationship exists under national law and practice.
For businesses, the practical consequence is that the classification of a workforce stops being a matter of what the contract says and becomes a matter of what the working arrangement actually looks like – who sets the price, who assigns the work, who monitors performance, who can sanction.
2. Algorithmic management
This is the part with the longest reach. Platforms must inform workers about automated monitoring and decision-making systems, may not process certain categories of personal data – such as data on emotional or psychological state, private conversations, or data used to infer trade union activity or beliefs – and must ensure human oversight of automated systems.
Crucially, significant decisions – restricting, suspending or terminating an account, refusing payment, or affecting a person’s contractual status – cannot be left to the algorithm alone. There must be human review and a route for the person to obtain an explanation and to contest the decision.
Where this connects to the AI Act
AI systems used for recruitment, task allocation, monitoring and evaluation of workers are classified as high-risk under the EU AI Act. A company using algorithmic management therefore sits at the intersection of three regimes at once: labour law, data protection and AI regulation. The obligations are different, but the evidence they demand is the same – documented logic, human oversight, testing, and records.
Practical example: a Swedish tech platform matching consultants
A platform matches freelance developers with client projects. It sets the rate range, ranks consultants by an internal score, allocates leads algorithmically and deactivates accounts that fall below a threshold. The consultants sign contracts as independent contractors.
Under the directive, most of the relevant indicators are present. The safest path is not to argue about the label, but to redesign the arrangement: give consultants genuine control over price and acceptance, make the ranking logic transparent, put a human in the loop before any deactivation, and give a written explanation and an appeal route when it happens. Those changes reduce misclassification risk and satisfy the algorithmic management rules at the same time.
Common mistakes companies make
Thinking it only applies to gig platforms. Any digital labour platform organising work through algorithmic systems is potentially in scope.
Relying on the contract. Written self-employment terms do not defeat a presumption based on facts.
Automating deactivation. An account termination decided by a score, with no human review and no explanation, is precisely what the directive prohibits.
Ignoring the data limits. Emotion analytics, private communications and inferences about union membership are off limits – and are also a GDPR problem.
Recommended actions
Inventory every automated system that monitors, ranks, allocates or sanctions people who work for you. Document what each system does and what data it uses. Introduce human review before any significant decision, and a documented appeals process. Delete or block data categories the directive prohibits. Review your contractor arrangements against the control indicators, and fix the ones that will not hold. And watch the Swedish implementing legislation – the detail of the presumption will be decided nationally.
Frequently asked questions
Does this apply to companies outside the EU?
The directive applies to platform work performed in the EU, regardless of where the platform is established.
Are the algorithmic management rules limited to employees?
No. They apply to persons performing platform work, including genuine self-employed people, which is what makes them so far-reaching.
What if Sweden is late with transposition?
The deadline binds the state, but companies should plan for the substance, not the delay. Late transposition usually means a shorter runway, not a lighter obligation.
Conclusion
The Platform Work Directive is the first EU instrument that regulates algorithmic management directly, and the first to shift the burden of proof on employment status onto the company. Businesses that use software to allocate and evaluate work should treat December 2026 as a design deadline, not a legal one – because the changes that satisfy it are changes to how the product works.
Lawgent advises companies on employment, AI and data governance where they overlap – algorithmic management reviews, worker classification, and AI Act readiness for HR and workforce systems. Book a free first hour and we will tell you where you stand.
