🚧

EnduraX™ is in pre-launch beta — for invited testers only. Features may change and the app is still a work in progress. By using this app you acknowledge it has not been officially launched.

AI Disclosure

How EnduraX uses AI — and what it won't do.

We use Anthropic's Claude models for three narrow features: athlete-side workout suggestions, coach-side training-plan modification suggestions, and the fueling meal estimator. Each feature sends different data, so this page describes each one separately — what it does, what data is sent, what guardrails apply, and how to turn it off.

Last updated 2026-08-04. Material changes will be reflected here and announced in-product.

What the AI is used for

Athlete-side workout suggestions. Each morning, EnduraX can suggest a single workout for that day. The suggestion is sized to your DARS sub-scores — composite readiness, training load, recovery, sleep, fueling, and RED-S protection. The AI receives only the computed scores; it does not receive your name, email, or other identifying information.

Coach-side training-plan modifications. When a coach uploads an athlete's training plan, EnduraX overlays each session with the athlete's readiness data. Where intensity and readiness don't align, the AI flags the session and suggests a modification — for example, reducing volume, reducing intensity, or converting to a recovery session. The coach reviews and decides; the AI never auto-applies a change.

Fueling meal estimator. When you log a meal in the opt-in Fueling module, the AI converts your meal description — or an optional meal photo — into an approximate nutrient estimate. Estimates are deliberately conservative (when uncertain, the AI estimates low rather than high), and they are approximations for wellness tracking, not clinical nutrition analysis. If the AI is unavailable, meals still log without an estimate.

That is the full scope. EnduraX does not use AI for diagnosis, treatment, prevention, prediction of injury, or any other clinical determination. EnduraX is a wellness tool, not a clinical or medical device.

The guardrails

Before any prompt reaches the AI, EnduraX applies hard server-side constraints based on documented evidence-based thresholds. These constraints sit upstream of the AI — the AI cannot override them.

  • DARS composite below 50: only easy or recovery suggestions are permitted, regardless of what the AI returns.
  • D4 fueling below 4: no tempo, threshold, or VO2 work suggested. This is the LEA (low energy availability) protective threshold.
  • D5 RED-S protection below 4: only easy or recovery suggestions are permitted. This threshold matches the coach-side alert threshold for elevated RED-S risk patterns.
  • Recovery Mode active: AI suggestions are not generated. The athlete's clinical pathway content takes precedence.
  • Coach-side load increase: AI suggestions for coaches can only reduce or maintain planned load. The AI cannot suggest a harder session than the coach already planned.
  • Output validation: EnduraX validates every AI response against the same guardrails. If the AI drifts from a constraint, the output is corrected to the safest matching template.
  • Fueling estimates never inflate: the meal estimator is instructed to estimate low when uncertain, and EnduraX independently clamps its output server-side so a recorded estimate can never exceed the AI's own stated plausible range.

What data is sent to the AI

Each feature sends only the data it needs. The two suggestion features send computed scores and workout context; the fueling meal estimator sends the meal content you choose to log. Specifically:

Sent to Anthropic — workout & training-plan suggestions

  • • DARS composite score (0–100) and label
  • • The five domain sub-scores (D1–D5, each 0–10)
  • • Athlete training context: goal, primary sport, recovery-mode status (when relevant)
  • • Workout context for coach-side suggestions: workout type, duration, target zone, RPE, notes
  • • The guardrail context (which intensities are permitted today)

Sent to Anthropic — fueling meal estimator

  • • The meal description you type when logging a meal
  • • The meal photo, if you choose to log by photo

Before sending, EnduraX screens meal text for identifiers (email addresses, phone numbers, dates, ID-like number runs) and blocks the AI call if any are found, and strips all photo metadata (GPS location, capture time, device info) — if metadata can't be stripped, the photo is not sent. These checks can't catch everything you might write or photograph, so avoid including names, faces, or other personal details in meal logs. Either way, the meal itself still logs.

NOT sent to Anthropic (any feature)

  • • Your name
  • • Your email address or other contact information
  • • Your account ID
  • • Raw biometric data (HRV, sleep stages, body weight, menstrual data)
  • • Personal notes you write in your check-in
  • • Health-modifier data (medications, hormonal contraception, thyroid status)
  • • Coach notes attached to athlete records

Anthropic processes this data under Anthropic's Commercial Terms of Service and Data Processing Addendum, which prohibit Anthropic from training models on our data. Under Anthropic's standard retention policy, API inputs and outputs are automatically deleted within 30 days; content flagged by Anthropic's safety systems may be retained longer, per their published policy. EnduraX does not store the prompts or responses in a long-term log accessible by employees.

Your control

Athletes: AI workout suggestions are surfaced when you ask for them — they are not pushed to you. You can use them, dismiss them, or ignore them entirely. Suggestions never auto-apply to your training calendar; you choose what to do with each one. To turn the feature off entirely for your account, visit Settings → AI Features and disable the toggle.

Coaches: AI training-plan modification suggestions require explicit opt-in before any AI feature runs against your athletes' data. You can opt in, opt out, or change your mind at any time from Settings. AI features are off by default. The opt-in is per-coach — every coach makes their own choice. If you opt in, every individual suggestion still requires your review before any change is applied to a training plan.

Fueling: the Fueling module is off until you opt in, and AI meal estimation runs only inside it — logging a meal is always your action; nothing is estimated in the background. You can turn the Fueling module off from Settings at any time (your data is kept), or delete your fueling data entirely.

Recovery Mode overrides everything. When an athlete is in Recovery Mode, AI suggestions are not generated. The athlete's clinical pathway library takes precedence, and the athlete's clinical care provider — not EnduraX — is the source of truth for their rehab.

Regulatory framing

EnduraX's AI features are designed and disclosed in line with the FTC's AI guidance (2023–2024 staff letters), GDPR Article 22 (right not to be subject to automated decision-making), and the Colorado AI Act (SB24-205) consumer disclosure requirements. We treat AI suggestions as decision-support tools for athletes and coaches — never as autonomous decisions.

EnduraX is a wellness product, not a medical device. The AI features do not diagnose conditions, prescribe treatment, predict injury, or replace clinical judgment. If your readiness data raises concerns, work with your healthcare provider, sports medicine clinician, sports dietitian, or athletic trainer.

EnduraX is subject to the FTC Health Breach Notification Rule (16 CFR Part 318). In the event of a breach of unsecured personal-health-record-related information, we will notify affected users, the FTC, and (where required) the media within 60 days as required by the rule.

Models we use

EnduraX uses two Anthropic Claude models, each scoped to specific features:

  • Claude Haiku 4.5 — used for athlete-side workout suggestions (scoped to single-day, single-athlete decisions) and for fueling meal estimates from text or photos.
  • Claude Sonnet 4.6 — used for coach-side training-plan modification suggestions. Scoped to per-workout decisions on a coach-uploaded training plan.

We may update models for performance or cost reasons. Material changes to the model selection will be reflected in this page and announced in-product.

Questions

For questions about this disclosure, AI-related data handling, or opt-out: