AI Notice

1. Planned functions

• Meal analysis — estimate foods, portions and nutrition from a photo for the user to check and edit.
• Training review — summarise selected workout and recovery records and identify patterns or questions to consider.
• Training questions — answer general fitness questions using only the account context the user chooses to provide.
• AI Max reports — optional plateau, programme, cross-pillar and annual deep dives with longer user-selected context.

AI will not silently create or change a meal, workout, target, profile field or progression rule. The user must review and confirm any proposed entry.

2. Who may use AI

The non-AI ledger and planned AI are restricted to people aged at least 13. Signup requires an exact date of birth and an affirmative declaration that it is accurate. Ordivum does not use document or third-party identity checks; this is a proportionate self-declaration, not proof of age. A confirmed under-13 date blocks that email from new registration until the declared thirteenth birthday. Payment does not prove age and a subscription alone will not unlock AI.

Accounts aged 13–17 receive the same conservative content restrictions as every account, plus age-aware prompts, safety filtering, monitoring, reporting and clear routes to qualified or emergency help. Ordivum will not infer adulthood from behaviour or allow a user to weaken those protections. A false date can defeat self-declaration, so suspected under-13 use is investigated and the account may be suspended or deleted.

3. AI and provider transparency

Every interactive AI surface will say from the first interaction that the response is generated by AI, not a human coach. The product labels are “Training assistant”, “AI training review” and “Photo meal estimate”. The proposed route is paid Amazon Bedrock from Ordivum's server, using Amazon Nova and Anthropic Claude model families selected by task complexity. Model names and routing may change before launch. No provider credential will be exposed to a browser or iOS app, and no free or consumer AI service will receive real user information.

Before activation, the incorporated company must accept and record the applicable AWS and Anthropic commercial terms, data-processing terms, subprocessors, retention settings, geographic routing and international-transfer safeguards. Anthropic's rules for products made available to minors also require age assurance, age-appropriate filtering, monitoring and reporting, child-safety prompts and resources. Activation remains technically blocked until those controls and the final verified retention and transfer position are issued.

4. Data and choice

AI will be off by default. Before the first use, an unticked and versioned choice will identify the provider, the selected purposes, the health and fitness data involved, international processing, retention, material risks and how to withdraw. This choice will be separate from the core ledger's health-data consent and from subscription acceptance.

A request will send only the prompt, sanitised image or ledger fields needed for that function. Email, display name and direct account identifiers will be excluded unless strictly necessary and newly disclosed. Separate toggles control training, nutrition, body and recovery context, including per-thread overrides. No customer content is used to train Ordivum or sent to provider feedback/training. Withdrawing AI consent immediately stops new processing and offers either stopping AI or stopping AI and deleting retained AI history, without disabling the non-AI ledger.

5. Health and safety boundary

AI output is a probabilistic estimate and can be incomplete, biased, unsuitable or wrong. It is general fitness information, not medical, dietetic, physiotherapy or mental-health advice; not a diagnosis or treatment; and not an emergency service. It must not determine whether an exercise, diet, supplement or target is safe for a particular person.

• Meal analysis cannot identify allergens reliably; check packaging and professional advice.
• AI must refuse diagnosis, treatment, medication, eating-disorder, self-harm and emergency instructions and direct the user to appropriate help.
• AI must not encourage extreme weight change, body shaming, concealment from a parent or professional, compulsive exercise, unsafe supplements or performance-enhancing drugs.
• High-impact or solely automated decisions about access, price, account status or another legal or similarly significant matter are prohibited.
• Outputs will be safety-filtered, tested across relevant users and use cases, monitored, and reportable from the same screen.

6. Accuracy and control

Meal calories, portions, exercise interpretations and training patterns will be presented as estimates with enough context for the user to correct them. Ordivum will record the model family and policy version needed to investigate a harmful or materially incorrect response, while avoiding unnecessary content retention. Sanitised meal images and provider payloads are transient; unconfirmed meal drafts last no more than 24 hours; synced AI threads/reports have a rolling 12-month limit. Material provider, purpose, data-category, retention or processing changes require a new assessment and, where needed, fresh consent.

7. Paid access

If AI is sold, the purchase screen will state the exact included allowances, price, billing interval, renewal and cancellation terms before payment. Ordivum will not create an unannounced usage charge or bill an overage without express agreement. Service-side spending and per-account limits may pause AI to protect the service, but will not block access to the non-AI ledger or data export. See the prepared subscription terms.

8. Rights, complaints and contact

The rights and complaint route in the Privacy Notice apply to AI processing. Users will be able to access or export retained AI records, correct the underlying ledger, delete retained AI history, withdraw AI consent and ask for a response to a concern. Contact thomaswdelaney115@hotmail.com.