A new policy framework is giving European employers a practical way to decide what workers should know about artificial intelligence. Published by Interface on August 25, the proposal divides workplace AI literacy into three cumulative levels based on how people use AI and the consequences of mistakes. It arrives as companies face a binding European Union duty but no universal test of competence.
- Framework
- Three cumulative workplace AI literacy tiers
- Baseline
- Relevant staff, contractors and service providers
- Operational
- Workers required to use AI systems
- Oversight
- People operating or overseeing high risk AI systems
Three tiers follow risk, not job title
The Interface framework starts with a baseline level for relevant staff, contractors and service providers. Workers at this level should understand which AI systems an organisation uses, the possibility of fabricated outputs or bias, risks to data and privacy, and where to raise concerns.
The second level covers people who use AI systems in their jobs. Their training would go further into system limits, verification of outputs, effective prompting, disclosure duties, copyright, escalation routes, and the legal and ethical rules connected to their work.
The third level is reserved for people who operate or oversee systems classified as high risk. Interface says they need deeper competence in monitoring, detecting bias or malfunction, intervening when necessary, and distinguishing high risk uses from prohibited ones. The level follows the worker's interaction with AI, not seniority or technical background.
The law requires action but not a universal score
Article 4 of the EU AI Act has applied since February 2025. The revised rule requires providers and deployers of AI systems to take measures that support AI literacy among staff and other people using those systems on their behalf.
The European Commission's guidance says employers must consider workers' knowledge, experience and training, along with the context and risk of each system. It also confirms that organisations do not have to guarantee a particular level of knowledge, issue certificates, or follow one mandatory course format.
That flexibility reaches ordinary businesses, not only companies that build AI. The Commission says an employer whose staff use ChatGPT to write advertising text or translate material should inform them about specific risks such as hallucinations. Stronger training requirements remain for staff responsible for human oversight of high risk systems.
Training records are not proof of competence
EU Perspectives reported on August 31 that companies still lack a consistent way to show what workers have learned. Interface researcher Catherine Schneider said many programmes concentrate on adoption and productivity while giving less attention to ethics.
Interface recommends documenting training and the reasons it matches each role, but it argues that measurement should come before certification. A course completion record can show that an employer acted. It cannot by itself show that a worker can recognise a false output, protect sensitive data, or intervene when an AI system creates risk.
The three level model is not an official EU standard. It is a policy proposal intended to make a broad duty easier to apply and compare. Its value will depend on whether employers test understanding and adapt training when systems, responsibilities or risks change.
Primary source
Check the original source
European Commission AI Literacy Questions and Answers is the source to consult for the underlying data, statement, ruling or live context.