EU AI Act compliance for proptech SaaS

PropTech SaaS using AI for tenant screening, rental decisions, mortgage assistance, or housing eligibility assessment sits on Annex III point 5(d) of the EU AI Act as a high-risk system. The Chapter III obligations apply on 2 August 2026: conformity assessment, training data governance, demographic bias testing, human oversight, accuracy reporting, post-market monitoring. Penalty ceiling is €15M or 3% of global turnover. Property-management portfolios and large landlords are now requiring AI Act attestations in their tech-vendor reviews.

Is your product high-risk under Annex III?

Annex III point 5(d) lists AI used in evaluation of natural persons in essential public and private services as high-risk. For proptech, the relevant use cases:

  • AI for tenant screening and rental application assessment
  • AI for mortgage and housing-loan creditworthiness (overlap with point 5(b) credit scoring)
  • AI for public-housing eligibility evaluation
  • AI for property-management workforce scheduling that involves residents

General-purpose proptech features (property listings, search ranking, market analytics, virtual staging, AI-generated listing copy) are NOT typically in Annex III scope. The line is whether the AI affects access to housing for specific natural persons.

Article 50 transparency obligations

Article 50 applies on top:

Article 50(1): every applicant interacting with your AI must be told they are talking to AI. Concierge chatbots, intake assistants, and viewing-booking tools all require disclosure.

Article 50(2): AI-generated listing copy, AI-generated property descriptions, AI-staged photographs, and AI-summarised tenant references must be marked as AI-generated.

Article 50(4): AI-generated virtual tours, AI-modified property imagery, and AI-synthesised landlord communications must disclose synthetic origin.

GDPR Article 22 overlap: tenant rejection decisions made by AI require a documented human-review path surfaced to the applicant.

Self-audit checklist before 2 August 2026

Seven checks before 2 August 2026:

  1. List every AI feature: listing generation, search ranking, tenant screening, application processing, fraud detection, market pricing, virtual staging, chatbot.
  2. Identify which features fall under Annex III point 5(d). Tenant screening, application scoring, and creditworthiness assessment are clear in-scope.
  3. For high-risk features, document training data composition. Demographic bias testing is required under Article 10, particularly given historical discrimination patterns in housing.
  4. Build the Article 43 conformity assessment file.
  5. Implement landlord-facing override controls per Article 14. Document the override workflow.
  6. Add Article 50 disclosures to applicant-facing chatbots, intake forms, and rejection communications. Surface GDPR Article 22 human-review option in every automated rejection.
  7. Set up Article 73 incident reporting for discriminatory outcomes or systematic application failures.

Penalties and enforcement

Penalty ceilings under Article 99:

  • High-risk tenant-screening failures: €15M or 3% of global turnover
  • Article 50 failures: €15M or 3%
  • Article 10 bias-testing failures: €15M or 3%

Worked example: a proptech SaaS with €3M ARR faces a theoretical maximum of €90,000 per violation under SME proportionality. Bigger cost: discrimination claims under EU Race Equality Directive 2000/43 or national housing laws. Documented housing-discrimination AI cases have resulted in fines of €500,000 to €3M plus mandatory remediation orders that effectively shut down the product line until fixed.

Last updated: 2026-05-28