FreyaSign in
GDPR & EU AI Act

Trust & law at Freya

Freya is developed and operated by NI – Negotiation Intelligence AG. It matters to us that you know exactly how we handle your data and where artificial intelligence is used in Freya. This page summarises that transparently.

Data protection (GDPR)

Controller. The controller for the data processing is NI – Negotiation Intelligence AG, Lucerne (UID CHE-246.633.079). For questions, contact us at privacy@negotiationintelligence.com.

Legal bases. We process your data on the following legal grounds:

  • Art. 6(1)(b) GDPR – to run your training sessions and provide your feedback (contract performance)
  • Art. 6(1)(a) GDPR – for analytics and tracking, only with your consent via our cookie banner
  • Art. 6(1)(f) GDPR – for security and product improvement (legitimate interest)

Your rights – in your hands. Under /settings/account you can at any time:

  • export all your data as a JSON archive (profile, sessions, transcripts, etc.)
  • request deletion of your account (deletion after a 30-day grace period)

Retention. Session transcripts and AI logs are deleted after 365 days. Aggregated progress data (e.g. your NELO rating, achievements) is retained beyond that so you can track your development.

Security. Details on our technical and organisational security are in our Trust Center, including:

  • Data primarily hosted in the EU (our Lovable AI Gateway is EU-routed); individual AI sub-processors process data in the EU and the US
  • Encryption with TLS 1.2+ and AES-256
  • Commitment to report data incidents within 72 hours
  • Overview of our sub-processors

International data transfers. Where data is transferred to the US (e.g. to OpenAI or ElevenLabs), this happens on the basis of Data Processing Addenda (DPAs) that include the EU Commission's Standard Contractual Clauses (SCCs) for EU→US transfers. These contracts are in force and linked in the Trust Center:

  • OpenAI DPA (incl. SCCs)
  • ElevenLabs DPA
  • Lovable AI Gateway DPA (as sub-processor)

OpenAI contractually guarantees that no training data for their own models is derived from our API requests ("no training on API data").

Use of artificial intelligence

Freya uses AI to give you realistic negotiation simulations and individual feedback. Specifically:

  • Language model for simulation & feedback: Google Gemini Flash for simulating your AI negotiation partner and for analysis and feedback on your negotiation performance
  • Voice synthesis: ElevenLabs for the voice of your AI negotiation partner
  • Knowledge-base retrieval: OpenAI is used exclusively for embeddings and internal retrieval from our knowledge base – not for your conversation with the AI partner and not for your session feedback
  • Knowledge base: Feedback and scoring are based on our proprietary negotiation knowledge base, not on general internet knowledge

By default, Freya does not automatically continue to learn from your personal data – you can actively enable this option in your user profile if you would like your training data to contribute to improving your individual training.

Labelling under the EU AI Act

We label AI-generated content clearly and transparently, as required by the EU AI Act (Art. 50):

  • You are informed at the start of every session that your counterpart is an AI simulation – not a real person
  • Feedback, scoring and evaluations are labelled as AI-generated

Classification under the EU AI Act

Freya is a training and practice tool for personal skill development. It is not a high-risk system in the sense of the AI Act, because:

  • Freya is used solely for self-training and does not make or prepare automated HR decisions (hiring, promotion, bonus, termination)
  • in a team/company context, managers only see aggregated participation figures, never individual results
  • Freya is not used as an assessment or selection tool for candidates
  • no biometric emotion recognition takes place

Freya therefore falls into the "limited risk" category with the corresponding transparency obligation, which we meet – as described above – through clear AI labelling.

Human oversight. In the Premium and Pro tiers, you can additionally have your AI feedback reviewed manually by a human. Fundamental concerns or anomalies are escalated directly to our NI team.

Simulated personality profile (DISC)

The only "profiling" in Freya concerns not you, but your AI training partner: before each session you select one of 16 simulated DISC personality profiles for your virtual counterpart. These are purely fictional, simulated profiles – no profiling of real people takes place.

Note: A future B2B feature called "Counterpart Profiling" is planned, but not currently implemented. No data processing takes place for it at this time.

This page is updated on an ongoing basis as our processes, technologies or legal classifications change.

Visit our Trust Center · FAQ