New guidelines on prohibited AI practices
The Danish Agency for Digital Government has published new guidelines on prohibited AI practices. The guidelines focus on areas under the Agency’s supervisory authority, including AI systems that infer emotions in the workplace, exploit vulnerabilities, or employ manipulative practices.
The EU AI Act applies to different AI solutions, actors and affected individuals. The term "AI solutions" covers AI models and AI systems. AI models can be an underlying algorithm. AI systems will have a broader user interface and perform a function with an AI model integrated. This could, for example, be a chatbot.
The new guidelines cover five types of prohibited AI practices:
- Harmful manipulation and deception
- Emotion recognition in workplaces and educational institutions
- Exploitation of vulnerable individuals and groups
- Social scoring
- Non-targeted scraping of facial images
The Agency is responsible for supervising these areas.
IUNO’s opinion
Even though the rules on prohibited AI practices have been in effect since 2 February 2025, the new guidelines are a helpful resource for companies still getting to know the new rules.
From an employer’s perspective, the guidelines on AI that detects emotions in the workplace are particularly interesting. More companies are using new methods, especially as part of the recruitment process. Such practices can become expensive if the system is prohibited.
IUNO recommends that companies be aware that systems capable of analysing facial expressions or tone of voice during job interviews are generally prohibited. The same goes for tools that try to assess an employee’s mood by analysing their voice in customer interactions.
[The Danish Agency for Digital Government’s guidelines on prohibited AI practices of 22 October 2025]
The EU AI Act applies to different AI solutions, actors and affected individuals. The term "AI solutions" covers AI models and AI systems. AI models can be an underlying algorithm. AI systems will have a broader user interface and perform a function with an AI model integrated. This could, for example, be a chatbot.
The new guidelines cover five types of prohibited AI practices:
- Harmful manipulation and deception
- Emotion recognition in workplaces and educational institutions
- Exploitation of vulnerable individuals and groups
- Social scoring
- Non-targeted scraping of facial images
The Agency is responsible for supervising these areas.
IUNO’s opinion
Even though the rules on prohibited AI practices have been in effect since 2 February 2025, the new guidelines are a helpful resource for companies still getting to know the new rules.
From an employer’s perspective, the guidelines on AI that detects emotions in the workplace are particularly interesting. More companies are using new methods, especially as part of the recruitment process. Such practices can become expensive if the system is prohibited.
IUNO recommends that companies be aware that systems capable of analysing facial expressions or tone of voice during job interviews are generally prohibited. The same goes for tools that try to assess an employee’s mood by analysing their voice in customer interactions.
[The Danish Agency for Digital Government’s guidelines on prohibited AI practices of 22 October 2025]
Similar
Managing transfer impact assessments in practice
Secret audio recordings of conversations led to serious criticism
Updated guidelines regarding breaches of data security
New rules on CCTV monitoring
The new NIS 2 Act has entered into force
New draft bill to protect against misuse of deepfakes