How we build Aury
Care you can check
Aury talks with people who are under pressure. That is a responsibility.
Here we describe who is behind Aury, how we assure its quality, where the limits of the AI lie and how we handle your data.
How we control what Aury does
The language model puts the words together. What Aury may and may not do is set by instructions our team writes, tests and approves.
Who is behind Aury
Aury is built by a team of psychologists, therapists and engineers.
Clinical responsibility
The clinical work is led by Saskia Fester, psychologist and trained therapist, co-founder and Chief Clinical Officer.
We are advised on the science by Prof. Dr. Kevin Hilbert (clinical psychology) and Prof. Dr. Christoph Benzmüller (AI systems).
People behind the AI
- Aury follows instructions our team writes, tests and approves. The language model does the wording; we set the boundaries.
- Our psychologists choose the methods: cognitive behavioural therapy, ACT, systemic counselling and mindfulness. Our engineers build and test the technology.
- Our psychology team regularly reads individual conversations afterwards to make sure Aury responds safely and professionally. Only our clinical team does this, and only within the scope of your consent.
Our clinical quality assurance – and the limits of AI
Aury is an AI coach for everyday stress, with safety built into its architecture from the start. Aury does not replace human support, psychotherapy or psychiatric treatment.
Clear limits
- Aury is for adults aged 18 and over. Before the course ‘Digital Stress Management with Aury’, you confirm a short health check.
- In the background of every conversation, an automated safety system looks for signs of a crisis. If something sounds worrying, Aury pauses and gently asks. If there are signs that you need professional help from people, Aury says so clearly and helps you find the right numbers and places to turn to.
- Important: no one reads along live, and Aury is not a crisis service. In an acute crisis, please contact emergency services directly (112).
Clinically developed, continuously tested
- Our clinical psychologists have developed their own benchmark of 15 quality criteria that we use to measure Aury's ethical and professional quality. Clinical psychologists rate Aury's behaviour in conversations against this benchmark.
- In addition, more than 100 automated tests check Aury against the same benchmark (‘LLM as a judge’: a second AI model rates the answers). Every problem we find becomes a permanent test and feeds into product development.
Your data
Aury Care GmbH is certified to ISO/IEC 27001:2022, the international standard for information security.
Security and data protection
- Aury is hosted in the EU. The AI models run in the EU as well. Your data is encrypted in transit and at rest.
- Your conversations are not used to train AI models.
- You can delete your account in the app at any time.
- Questions about your data? Our data protection contact answers them: dpo@aury.co.
Certificates and publications
- ISO/IEC 27001:2022 – certificate DE-IS-20260240, valid until February 2029
- “Sicherheit der Nutzenden in der psychologischen KI-Beratung” – Fester & Hilbert, Verhaltenstherapie & psychosoziale Praxis 57(3), pp. 408–423, 2025
- “Einsatz von KI bei psychischen Belastungen – eine Reflexion” – Fester, Rank & Lägel, Market Access & Health Policy 05/2026, pp. 36–39
- “Script-Based Dialog Policy Planning for LLM-Powered Conversational Agents” – Wasenmüller, Hilbert & Benzmüller, arXiv:2412.15242, 2024
- Pre-registered study with Universität Osnabrück – osf.io/a7q8r
- Trust Center with the certificate, scope and controls
Scientific partners: Osnabrück University, University of Bamberg, HMU Erfurt
What is not finished yet
Whether our course ‘Digital Stress Management with Aury’ reduces stress is being studied in the pre-registered study (see above). There is no proof of effectiveness yet – until there is, we do not claim it.
Aury is new, in its testing phase, and makes mistakes. We don't wait for something to go wrong: we keep watching (see the review processes above) and keep improving Aury. Feedback is always welcome 💜
As of September 2026