AI Transparency Policy

CAMARON uses artificial intelligence across 20+ aquaculture decision-support capabilities, organized into a 6-module AI Suite plus integrated intelligence features: • Disease Detection: Analyzes uploaded images and symptom descriptions to identify potential diseases, matching against 143+ disease protocols and 163+ species profiles. • Feed Optimization & Yield Coach: Recommends feed formulations, feeding rates, and schedules based on species, life stage, water parameters, and growth targets. • Growth Forecasting & Market Forecaster: Projects harvest dates, biomass estimates, revenue, and price trends across 26+ export markets. • Water Quality & Pond Doctor Chat: Interprets sensor readings against species-specific optimal ranges and answers operational questions in 31 languages. • Compliance Reviewer & Claim Assistant: Audits documentation against BAP/ASC/GlobalG.A.P. requirements; drafts insurance claim narratives. • Negotiator & Risk Scoring: Suggests counter-offers for input procurement and aggregates farm risk into lender-ready scores. • Vision Pipeline & Delta-T Analysis: Photo-diagnoses creature health and evaluates air-water temperature differentials for stocking/harvesting windows. All AI modules use Google Gemini models via the Lovable AI Gateway. No user data is sent to third-party AI providers in identifiable form.

Each AI analysis follows a structured pipeline: 1. Data Collection: The system gathers relevant inputs — sensor readings, crop logs, uploaded images, user-provided symptoms, and species profiles from our reference database. 2. Context Enrichment: Inputs are enriched with aquaculture domain knowledge from our curated knowledge base (species profiles, disease protocols, water quality standards, feed formulations). 3. Model Processing: Enriched inputs are sent to the AI model with structured prompts that constrain outputs to aquaculture-relevant recommendations. 4. Output Validation: Model responses are parsed into structured formats (JSON) with confidence scores where applicable. 5. Presentation: Results are presented with clear explanations, confidence indicators, and references to supporting data. All AI outputs include the model used, processing latency, and are logged in our ai_analysis_logs table for auditability.

CAMARON's AI provides decision-support tools — not substitutes for professional veterinary, agricultural, or financial advice. • Disease detection accuracy varies by image quality, species, and disease type. Results should be confirmed by qualified aquaculture veterinarians. • Growth forecasts are statistical projections based on historical patterns and may not account for unprecedented environmental events. • Feed recommendations are optimized for common scenarios but may not address all nutritional requirements for every culture system. • Market intelligence reflects available data and trends; actual market conditions may differ. • Water quality predictions assume sensor calibration is current and accurate. We continuously improve our models using anonymized, aggregated data. Individual accuracy metrics are published in our quarterly transparency reports. Users should exercise independent judgment and consult domain experts for critical decisions.

Your data is used for AI in two ways: 1. Real-time Analysis: Your specific data (sensor readings, images, crop logs) is sent to the AI model during each analysis request. This data is not retained by the AI model provider after processing. 2. Model Improvement: Anonymized, aggregated data derived from platform usage may be used to improve our AI models and knowledge base. This includes: • Statistical patterns in water quality parameters across regions • Disease prevalence and treatment outcome data (fully de-identified) • Growth curve benchmarks by species and culture system You can opt out of contributing to model improvement via your account settings (Settings → Privacy → AI Data Contribution). Opting out does not affect your access to AI features. Individual user data is never used to train models — only aggregated, anonymized datasets.

CAMARON maintains human oversight of AI systems through: • Expert Review: All AI disease detection results flagged as high-severity are automatically routed to registered aquaculture doctors for review via our expert consultation system. • User Control: You can accept, modify, or reject any AI recommendation. AI outputs are suggestions — they do not trigger automated actions without your explicit confirmation. • Actuator Safety: IoT actuator commands (aerator control, feeding systems) require explicit user approval even when AI recommends action. Emergency overrides are always available. • Content Moderation: AI-generated content in the platform (forum responses, knowledge base) is monitored for accuracy and appropriateness. • Audit Trail: All AI analyses are logged with inputs, outputs, model version, and timestamps. You can review your AI interaction history in your account dashboard.

We commit to the following principles: • Transparency: We disclose which features use AI, what models power them, and how decisions are made. • Fairness: Our models are evaluated for bias across species, regions, and farm sizes. We do not discriminate based on user demographics. • Privacy: AI processing uses the minimum data necessary. Personally identifiable information is stripped before any aggregated analysis. • Safety: Critical aquaculture decisions (medication, chemical treatments, harvest timing) always include disclaimers and recommendations to consult experts. • Accountability: Our AI Ethics Committee (comprising aquaculture scientists, data ethicists, and engineering leads) reviews AI model updates quarterly. • Compliance: Our AI practices align with the EU AI Act (risk-based approach), India's emerging AI governance framework, and aquaculture certification standards (BAP, ASC, GlobalG.A.P.).

The following uses of CAMARON's AI outputs are prohibited: • Using AI disease detection results as the sole basis for administering restricted or prescription medications without veterinary consultation • Redistributing, reselling, or white-labeling AI outputs or analysis results • Using AI market intelligence to engage in price manipulation or market fraud • Submitting intentionally false data to manipulate AI model outputs • Attempting to extract, reverse-engineer, or replicate CAMARON's AI models, prompts, or training data • Using AI outputs to generate misleading certification or compliance claims Violations of these restrictions may result in account suspension or termination under our Acceptable Use Policy. For questions about permitted uses, contact ai-ethics@robicrufarm.com.