Ce qu'est Prism, et ce qu'il n'est pas.

Prism reads what you write and gives you back patterns:

·Classifications (topics, themes, emotional register).

·Semantic connections between entries.

·A continuously updated behavioural model of recurring patterns, strengths, and blind spots.

·A weekly reflection called the Mirror.

Those outputs are produced by large language models and embedding models running against your own writing. They are signals derived from your inputs. They are not a diagnosis, not a prediction of your future behaviour, and not advice.

·Not a licensed mental-health service. Under UAE Federal Law No. 10 of 2023 on Mental Health, “psychological counselling” and “psychiatric assessment” are reserved to licensed professionals working in licensed mental-health facilities. Prism is not such a facility, no Prism output is counselling or assessment within that statute, and no Prism employee or contractor performs licensed mental-health services through Prism. AI output is statistical and probabilistic pattern summaries of what you wrote — not clinical findings.

·Not a medical device. Prism is not registered with any medical regulator. Nothing it produces is a clinical assessment, diagnosis, screening result, or treatment recommendation.

·Not a substitute for mental-health care. If you are seeing a therapist, psychiatrist, or counsellor, keep seeing them. If you think you need one, find one. Prism does not replace that relationship.

·Not a financial adviser. Users sometimes journal about trading and investment decisions. Any pattern Prism surfaces about your trading psychology is not investment advice, not a market prediction, and not a recommendation to buy, sell, or hold anything. Do not make a financial decision because Prism said you behave a certain way under stress.

·Not a legal adviser. Prism does not provide legal advice and its outputs are not legal opinions.

·Not a lie detector, not a personality test, not an employer-useful assessment. The behavioural model is a reflective tool for the user, nothing else. Do not use Prism to assess other people, screen job candidates, or make decisions about anyone other than yourself.

Ne prenez aucune des décisions suivantes en vous basant uniquement sur la sortie de Prism. Celles-ci sont réservées au jugement humain, aux conseils professionnels agréés, ou aux deux :

·Medical and medication decisions. Starting, stopping, or changing dosage of any medication; self-diagnosis; deferring or declining medical care.

·Mental-health crisis response. Including deciding whether or not to seek emergency help. If in doubt, call 999 (UAE) or your local crisis line.

·Trading, investment, or financial decisions. Any transaction, allocation, or position sizing based on Prism’s read of your “trading psychology”.

·Relationship-termination or similarly consequential interpersonal decisions. Ending a relationship, a marriage, or a guardianship arrangement based primarily on Prism output.

·Self-harm decisions. Do not use Prism as a decision aid in any context involving self-harm or harm to others.

·Safety-critical conduct. Driving, operating machinery, or performing other tasks where impaired judgement causes risk, based on Prism’s read of your state.

Parce que la sortie de l'IA peut être avec assurance incorrecte, un humain doit exercer un jugement indépendant sur tout ce qui précède.

·Confidently wrong. Language models produce fluent output even when the underlying inference is mistaken.

·Overgeneralising. A single entry can nudge the behavioural model in ways that don’t reflect who you are across time.

·Missing context. The model sees only what you wrote; it doesn’t know what happened outside your writing.

·Mirroring your framing. If you describe yourself harshly, the model may reinforce the harsh framing. This is a known failure mode of reflection-style AI.

·Changing over time. Because we continuously evolve the pipeline, outputs one week may look different from the same input the previous week.

Traitez la sortie de l'IA comme un point de départ pour votre propre réflexion, non comme un verdict.

·Use them as raw material for your own thinking.

·Compare them to your lived experience; discard anything that doesn’t match.

·Take anything that does match to a licensed professional if it warrants that.

·Do not make irreversible decisions — clinical, financial, relational — based on a Mirror or behavioural-model output alone.

·Zero-retention AI processing. Journal entries are sent to Claude and OpenAI under their zero-retention API modes; vendors do not store them and do not train models on them. See Privacy Policy and Subprocessors. Politique de confidentialité Sous-traitants.

·No profiling for advertising. We do not feed your writing or inferences to any advertising system.

·No sale of inferences. We don’t sell the behavioural model, the classifications, or the Mirror to any third party.

·No automated decision with legal effects. Prism does not make automated decisions that have legal or similarly significant effects on you. If we ever introduce one, we’ll obtain fresh explicit consent first.

Our liability for any decision you make based on AI output is limited per the Terms of Service. You agree not to hold Prism liable for outcomes arising from your own decisions informed by the Mirror, the behavioural model, or any other AI signal produced by the service. Conditions d'utilisation.

This liability limitation does not exclude liability that cannot be excluded by law. In particular, under UAE Consumer Protection Law (Federal Law No. 15 of 2020) and its Executive Regulations (Cabinet Resolution No. 66 of 2023) and the unfair-terms rules contained in them, we cannot waive our statutory duties of accuracy, safety, and fitness for purpose — and we do not attempt to. Nothing in this page or our Terms overrides any non-waivable consumer right you have under UAE or your local law.

Because Prism collects and analyses sensitive psychological data, researchers, journalists, and academics sometimes ask whether they can study our users. They cannot. We do not provide identifiable user data, aggregate inferences, or sampled outputs to any research entity. If we ever pursue research using anonymised, aggregate-level patterns, we will obtain separate explicit consent first and describe the study scope in clear language.

·Safety, moderation, crisis-related concerns: support@prismlens.net

·Privacy, data-subject requests: privacy@prismlens.net · dpo@prismlens.net.