AI Transparency & Model Limitations Notice

  1. Role of AI
    Auctralytix uses third-party large language models to generate natural-language explanations from analytical data that has already been computed by deterministic application code. The intended architecture does not use the language model to calculate the TPO distribution, point of control, value area, initial balance, session boundaries, day type, structural detector levels or other structured analytical values.


  2. Structured values versus narrative
    Structured values rendered by the application originate from the analytical engine. AI narrative may repeat or discuss those values in prose. Because generative models can make transcription or interpretation errors, a number quoted in AI-generated narrative may be incorrect even where the underlying structured value is correct. If there is any conflict, the structured engine-generated value controls.


  3. No predictive AI
    The Service is not intended to use AI to predict future price direction, score the probability of a future move, recommend a transaction or create an entry, exit, target or stop-loss level.


  4. Automated content controls
    Auctralytix may use automated scanning, restricted vocabulary, retries, suppression and human review to reduce the risk of directional or advisory wording. No content-control system can guarantee that every possible formulation will be identified.


  5. Third-party model risk
    Third-party AI systems may change, become unavailable, behave differently after provider updates, introduce latency or produce unexpected output. We may switch models, suppress narrative or fall back to deterministic text where necessary.


  6. User obligations
    Do not treat AI narrative as investment advice. Independently verify any information material to a trading or investment decision. Never rely on an AI-generated statement merely because it is fluent, confident or specific.


  7. Data protection
    Personal-data processing associated with AI providers is governed by the Privacy Notice and internal processor controls. The intended analytical prompt should minimise user-identifying data.