Housing organisations are increasingly turning to digital tools and algorithms to cope with the growing flood of applications. In sought-after locations, a single advert can quickly attract several hundred enquiries. However, the use of automated systems carries significant legal risks, particularly with regard to the General Equal Treatment Act (AGG).
What is the AGG and why is it relevant to letting?
The AGG prohibits discrimination on the grounds of characteristics such as ethnic origin, gender, religion, age or disability. Housing organisations are bound by this when allocating flats. Breaches can prove costly: courts frequently award those affected damages amounting to two to three times the monthly rent. And this applies not only to decisions made by a person, but also to errors made by an algorithm.
Algorithms do not discriminate on a case-by-case basis, but systematically
That is the real problem. A human case worker can reflect on individual cases, weigh up the circumstances and show empathy. A poorly trained AI system, on the other hand, consistently repeats learnt patterns without exception. If historical award data disadvantages certain groups, the AI learns precisely that as the norm and continues to do so. Address data can act as a hidden indicator of origin, whilst language patterns can serve as a proxy for ethnicity. Both are prohibited under the AGG.
The Federal Anti-Discrimination Agency reported that it had dealt with around 500 cases relating to the housing market in 2025, an increase of 25 per cent on the previous year. The actual number of unreported cases is likely to be significantly higher.
What landlords and housing associations need to bear in mind in practical terms
Carolin Klein, a solicitor at the law firm Heuking, sets out five requirements for allocating accommodation in accordance with the General Equal Treatment Act (AGG) using digital systems:
- Transparency of the selection criteria:It must be possible at all times to understand and explain who is assessed and on the basis of which criteria.
- No indirect discrimination:Seemingly neutral characteristics, such as an address or speech patterns, must not be used as a substitute for protected characteristics.
- Accurate training data:Systems based on distorted historical data perpetuate past injustices and are legally problematic.
- Random selection where candidates have the same qualifications:If several candidates are equally suitable, a draw must be held.
- Responsibility remains with the company:Regular spot checks and the analysis of complaints are mandatory, regardless of which software is used.
The EU AI Act requires human final decision-making
Fully automated rejections – for example, based on a credit score – are not permitted under the EU AI Act in the context of high-risk data processing. The final decision on a tenancy agreement must be made by a human being. Housing organisations that ignore this risk breaching not only the AGG but also European law.
What does this mean for private landlords and smaller companies?
Even those who let only a few flats and do not use complex software are bound by the AGG. Reasons for rejection should be documented in a way that is transparent. Anyone using a service provider or a letting platform is still personally liable for discriminatory letting decisions made through that channel. Where in doubt, it is advisable to seek legal advice before introducing new digital processes.
Anyone who wishes to check now whether their recruitment processes comply with the AGG should not put it off. Undocumented procedures in particular – those based on past experience or gut feeling – are virtually impossible to defend in court.
Source: Haufe: Property
This post and the accompanying image were created with the help of AI and have been editorially reviewed.