Skip to content
LEGAL & TRUST

AI disclosure

Homa uses AI to interpret bounded context and support reflection. It is not an autonomous trader, financial adviser, or authoritative source of market truth.

DRAFT — COUNSEL REVIEW REQUIRED

These documents reflect the intended Homa product and current marketing prototype. They are not effective terms and are not legal advice. Company identity, launch jurisdictions, provider terms, regulatory perimeter, and consumer-law treatment must be approved before production launch.

Effective date
Not yet effective — pending company details and qualified counsel approval
Last updated
16 August 2026
Version
0.5-marketing-draft

1. Where AI is intended to help

AI may support voice and text mentoring, multimodal analysis of a selected chart or region, methodology-aware explanations, non-clinical behavior classification, session summaries, and evidence-based reflection.

2. What context AI may receive

Depending on the feature and your settings, a model may receive your prompt or transcript, one authorized chart snapshot or crop, platform and timeframe context, selected methodology version, bounded session observations, prior relevant patterns, and server-computed rule evidence.

3. Separation of trust

System instructions, approved methodology, user input, screen-derived text, tool results, and stored memory should remain clearly separated. Text found inside a shared chart or window is untrusted and cannot override security policy, authorization, or tool permissions.

4. What AI must not do

  • Place, change, cancel, or close a trade.
  • Issue guaranteed buy, sell, or hold instructions.
  • Promise profit, certainty, win rates, or a guaranteed edge.
  • Present behavior analysis as a clinical diagnosis.
  • Fabricate precision when the visual context is unclear.
  • Override ownership checks, consent, or security controls.

5. Known limitations

Models can hallucinate, miss visual details, misunderstand methodology, lose conversational context, reflect bias, or produce inconsistent answers. Rules can also contain defects or encode a disputed interpretation. Stronger models and guardrails reduce but do not remove these risks.

6. Human control

You may ignore, correct, or reject AI output. Live mentoring should remain calm and suggestive rather than directive. Draft reflections and stored patterns should be editable and deletable where appropriate. Important decisions must be independently verified.

7. Model routing and provider controls

The intended backend should keep provider credentials and model routing server-side, use short-lived authorization for realtime voice, limit tools independently, meter cost, and allow AI, voice, visual analysis, behavior analysis, or new sessions to be disabled through server controls.

8. Regulatory boundary

Regulatory and counsel review required. If Homa behavior becomes personalised advice, a signal service, portfolio management, or execution, the feature must pause for regulatory review and any required licensing. Calling output educational does not determine its legal character.