ARC Intelligence – Data & AI Policy

Last Updated: 27 August 2026

This document explains how ARC Intelligence uses artificial intelligence and biomechanics technology, the nature and limitations of coaching content on the platform, and the responsibilities of users. It should be read alongside the Terms of Service, Privacy Policy, and AI, Biomechanics & Coaching Disclaimer.

1. Nature of ARC Intelligence

ARC Intelligence is AI that develops with the athlete: coaching, performance, and educational software. ARC Intelligence builds a persistent, evolving understanding of authorised athlete context and provides biomechanical video analysis as evidence, AI-generated performance insights, ARC Intelligence chat reasoning, Digital Twin (in the ARC app for eligible athletes), and a structured contributor ecosystem where coaches submit coaching insights through the Verified publishing system.

ARC Intelligence is not:

2. How Data Is Processed

ARC Intelligence uses a combination of on-device and secure cloud-based processing to deliver its services. We do not claim that all AI processing is self-hosted or that processing is “fully private” in a technical sense that excludes cloud providers.

2.1 On-Device Processing

Certain analysis features use Apple platform frameworks (CoreML, Vision) to perform pose estimation, movement tracking, and shot classification directly on your device. When processing occurs on-device, raw video frames are processed locally. However, derived results, metadata, and analytical outputs may be transmitted to ARC Intelligence servers for storage, further processing, or display.

2.2 Cloud-Based Processing

Videos and other content may be uploaded to secure cloud infrastructure (Supabase Storage) for contributor submissions, Verified content, and analysis workflows. ARC Intelligence uses third-party AI and machine learning providers to support analysis, moderation, insight generation, chat reasoning, Digital Twin synthesis, embeddings/retrieval, and related platform features. As currently configured in production code, the primary cloud LLM/embeddings provider is OpenAI (including OpenAI-compatible endpoints where configured for specific models). Named references to Anthropic or xAI in older materials should not be read as a current production dependency unless and until those providers are actually wired and disclosed again.

Data processed by these providers may include uploaded content, biomechanical measurements, generated insights, chat text, Athlete Understanding Context packets, Digital Twin synthesis inputs/outputs, metadata, and related operational information depending on feature usage. Separately, ARC Sports Agent may retrieve Google Workspace or Microsoft mailbox/calendar/file data to fulfill authorized user tasks. Sports Agent retrieval and execution paths do not themselves call OpenAI. As currently implemented, Gmail and Google Drive content retrieved by Sports Agent is not automatically added to the Athlete Understanding Context used for general ARC Intelligence conversations. If Google-derived information becomes part of an ARC Intelligence conversation, that conversation text may be processed by our configured LLM provider solely to generate that response, as described in the Privacy Policy §2.15. We do not use Google user data obtained via Google APIs to create, train, or improve a machine learning or artificial intelligence model beyond that specific user’s personalized model for the appropriate use case or user-facing feature (including MovementNet).

2.3 Audio Event Processing

Where applicable, audio event metadata extracted from content may be processed using cloud-based AI services for event detection and classification.

3. AI-Generated Analysis

ARC Intelligence uses artificial intelligence and machine learning (including proprietary MovementNet technology and third-party AI/ML providers) to generate biomechanical analysis, movement pattern recognition, performance scoring, coaching insights, chat responses, and Digital Twin content.

Important limitations:

3A. ARC Intelligence, Athlete Understanding Context, and Digital Twin

ARC Intelligence reasons over the athlete’s authorised context to generate recommendations, summaries, and interpretations. Authorised context may include Athlete Understanding Context packets (for example goals, milestones, retrieval extracts, prior analysis summaries, and permissioned profile or performance context). Organisation-wide ARC Intelligence, where available, is for the Organisation Owner (or other expressly granted contexts), not for every organisation member by default.

Digital Twin is the evolving summary and visualisation of ARC Intelligence’s current understanding of the athlete for eligible athletes in the ARC app, built from authorised evidence, conversations, goals and development over time. It belongs to the athlete and is not organisation-wide. It is not a biological simulation, medical model, or independent intelligence separate from ARC Intelligence. It may include inferred statements that are not fully athlete-confirmed. User corrections can improve stored understanding. Source deletion, access revocation, and account deletion affect what Twin and ARC Intelligence may retrieve going forward, subject to retention rules in the Privacy Policy.

Camera measurements and any illustrative image-generation outputs (where used) are separate. Some features use external model providers under their contracts and our configuration. Availability of Digital Twin and full ARC Intelligence chat is primarily in the iOS app; the website does not host those athlete product surfaces.

4. MovementNet & Proprietary Technology

MovementNet is ARC Intelligence's proprietary biomechanics technology encompassing structured analytical frameworks, movement taxonomies, scoring systems, and datasets built from large-scale movement analysis. MovementNet is continuously refined as additional data is analyzed.

ARC Intelligence also uses licensed open-source and commercial third-party AI models. ARC Intelligence does not own those underlying models. ARC Intelligence does own its proprietary datasets, biomechanics frameworks, structured insight systems, tagging taxonomies, MovementNet data, and derived analytical outputs.

5. Contributor-Generated Coaching Insights

The Verified publishing system allows approved coaches and contributors to submit structured coaching insights, skill area analyses, and annotations. These contributions are moderated for quality but come with the following limitations:

6. Separation of Content Types

Users should understand the source and limitations of each content type on ARC:

7. Sport Availability

Tennis is currently the primary sport with full video analysis and AI-powered insight capabilities on ARC. Other sports are available for profile registration, interest indication, and early-access enrollment. Analysis features for additional sports are under development.

Users should not rely on ARC Intelligence for analysis of sports that are not yet analysis-ready. The availability of sport selection during account creation does not imply that full analysis features are available for all listed sports.

8. Physical Activity Risks

All physical activity carries inherent risk of injury. Users of ARC Intelligence acknowledge and accept that:

9. No Professional Relationship

Use of ARC Intelligence does not create a coach-athlete relationship, medical provider-patient relationship, or any professional advisory relationship between ARC and the user, or between contributors and users. ARC Intelligence is a technology platform that provides tools and information. The responsibility for how that information is used rests with the user.

10. Data Retention for Analysis

Anonymized and de-identified biomechanical data (including pose data, movement metrics, and analytical outputs) may be retained indefinitely to support ongoing research, model training, and platform improvement. See the Privacy Policy for complete data retention details.

11. Limitation of Liability

To the maximum extent permitted by applicable law, ARC, its officers, employees, and contributors shall not be liable for any injury, loss, damage, or adverse outcome arising from reliance on AI-generated analysis, contributor-generated coaching content, training decisions made based on ARC Intelligence content, inaccuracies in biomechanical analysis, or any physical activity undertaken in connection with ARC Intelligence content.

See the Terms of Service for the complete limitation of liability.

12. Contact

For questions about AI analysis, data processing, or this disclaimer: hello@arcintelligence.online.