Mira (Psy) Addendum

This addendum applies when you use Mira — the NAL psychology companion at mira.nal.digital, the Eliza real-time copilot available to verified therapists, the psy-agents API, the edu-platform (therapist training), or any related Service that delivers mental-health-adjacent content.

It is part of the Terms of Service and incorporates the AUP, Privacy Policy, and Content License.

Effective date: 2026-07-14 (v1.0.1; originally published 2026-05-15)

⚠️ Important Notice

Mira is not a substitute for professional medical or mental-health care. Mira is not a crisis service. Mira is not a medical device.

If you are experiencing a mental-health emergency, suicidal ideation, self-harm urges, or a crisis of any kind, contact your local emergency number or a crisis line immediately:

• International association of suicide-prevention lines: https://findahelpline.com
• United States: 988
• United Kingdom: Samaritans 116 123
• Russia: 8-800-2000-122 (children), 8-495-989-50-50 (adults)
• European Union: 116 123

Mira may detect crisis language and surface this notice in-product, but you must not rely on Mira to recognise a crisis. The decision to seek help in a crisis is yours.

1. What Mira Is

1.1. Mira is an AI-powered mental-wellness companion that provides:

• structured journalling, CBT-style exercises, and ACT-style practices;
• a conversational interface for reflective practice;
• educational content drawn from published evidence-based frameworks;
• optional integration with a verified therapist if you choose to share with one.

1.2. Mira does not:

• diagnose mental-health conditions;
• prescribe or recommend medication;
• replace a licensed therapist, psychiatrist, or other clinician;
• provide medical advice;
• guarantee any particular wellness outcome.

2. Eligibility

2.1. You must be at least 18 years old to use Mira. Mira is not suitable for children. If you are between 16 and 18, you may use Mira only with active involvement of a parent or guardian — and at our discretion we may decline access.

2.2. Mira is not appropriate for users with active psychosis, mania, severe depression with active suicidal intent, active substance-use crisis, or other conditions that require clinical care. If any of these apply to you, please consult a clinician before using Mira.

3. How Mira Works

3.1. Your input to Mira (text, voice, journal entries, exercise responses) is processed by NAL's psy-agents service. AI inference is routed through the multi-agent proxy to upstream LLM providers (OpenAI, Anthropic, Google). The model used and its provider are recorded with each interaction.

3.2. Modes. Mira operates in several modes (Patient, Mentor, Supervisor, Eliza Live). Each has different prompts and guardrails; all share the disclaimer above.

3.3. Crisis triggers. Mira applies semantic crisis detection (not keyword laundry lists). When a possible crisis is detected, Mira:

• surfaces the crisis-line notice (Section "Important Notice" above);
• offers to connect you with a verified therapist if you have linked one;
• pauses any active exercise or session;
• does not call emergency services on your behalf — that decision remains yours.

3.4. Coach-mode guards. Mira applies session guards (loop detection, stall detection, prompt-injection mitigation on therapist notes). These are platform safety controls, not therapeutic interventions.

4. Privacy and Confidentiality

4.1. Mira sessions are private. They are not shared with other users, with NAL staff, or with researchers, except:

• with you, on request, in machine-readable form;
• with a therapist you have explicitly linked, and only within the scope you authorise;
• with a trained NAL safety reviewer, and only if you have enabled Safety Review Consent — an optional, off-by-default, revocable consent for human review solely to detect and prevent serious harm;
• where required by valid legal process or where NAL believes in good faith disclosure is necessary to prevent imminent harm to life (e.g., a credible threat of suicide or harm to another);
• as aggregated, de-identified statistics for service improvement (number of sessions, error rates) — never the content.

4.2. Mira sessions are not used to train AI models.

4.3. Clinical record retention. Where local law requires retention of clinical records (e.g., when a verified therapist conducts a session through Eliza), NAL retains the relevant records for the period required by law. The retention period is recorded on the session.

4.4. Voice recordings are transcribed and the transcript is retained as part of the session record. The original audio is retained for 30 days for quality review and is then deleted unless you have explicitly opted to keep it.

5. Therapist Tier

5.1. Some accounts are verified as therapists after submitting credentials through the therapist-application flow. Verification:

• requires uploaded government-issued ID and proof of professional qualification;
• is reviewed by an NAL administrator;
• grants access to Eliza Live, the therapist-supervisor view, and the ability to link with consenting client accounts.

5.2. Therapist duty of care. Verified therapists agree:

• to provide care within the scope of their licence and competence;
• to maintain clinical records as required by their licensing jurisdiction;
• to act on a duty to warn or to safeguard a client in line with their professional code;
• not to use AI features to substitute for clinical judgement that their licence requires them to exercise personally;
• to clearly identify themselves as the therapist in any therapeutic communication — no AI-only impersonation of a therapist;
• to maintain confidentiality of client material consistent with their professional code and applicable law.

5.3. Misrepresentation. Holding out as a verified NAL therapist without verification, or after a verification has been revoked, is a serious violation. NAL may report misrepresentation to relevant regulators.

6. AI Output in Therapeutic Contexts

6.1. AI suggestions inside Eliza and the supervisor view are decision support, not clinical advice. The therapist remains the responsible clinician.

6.2. Where Mira provides educational content drawn from evidence-based frameworks (HiTOP, CBT, DBT, ACT, etc.), the content is summarised and synthesised by AI. Synthesis may contain inaccuracies.

6.3. Mira does not produce diagnostic statements presented as authoritative. Where Mira identifies symptom patterns it does so to help you reflect, not to label.

7. Pricing and Wallet

7.1. Mira sessions may be priced per session, by subscription, or included free up to a daily quota. The price applicable is shown in-product before each paid action.

7.2. Free sessions. New users may receive complimentary sessions as part of onboarding. Promotional credits may expire — the expiry is shown when granted.

7.3. Refunds. Sessions that fail to complete (LLM error, network fault) are refunded automatically. For other refund requests see the Payments and Refunds Policy.

8. Education and Credentialing

8.1. The edu-platform offers training modules and assessments. Passing an edu-platform module is not a professional credential. It does not authorise the holder to practise psychology or psychotherapy in any jurisdiction.

8.2. The credit signal earned by completing modules can contribute to NAL-internal recognition (therapist applications, mentor roles) only.

9. Limits and Liability — Reinforcement

The limitations of liability in Section 13 of the Terms of Service apply with particular force to Mira. To the maximum extent permitted by law, NAL is not liable for:

• mental-health outcomes, positive or negative, that follow from your use of Mira;
• a failure of Mira to detect a crisis or to deliver content at a particular moment;
• decisions you make in reliance on Mira output.

Where applicable consumer-protection or medical-device law gives you non-waivable rights, those rights apply.

10. Reporting Concerns