AI Usage Policy

1. How Alfred Uses AI

1.1 Automatic Model Routing

Alfred does not use a single AI model. Instead, it automatically routes each request to the most appropriate model based on:

• Query complexity: Simple queries use efficient, cost-effective models; complex reasoning uses more capable models
• Task type: Coding tasks, image generation, web search, and transcription each use specialized models
• Your subscription tier: Higher tiers unlock premium image models and additional capabilities such as video generation and lipsync
• Cost efficiency: We route to the minimum capable model needed to deliver high-quality results

1.2 What Alfred Can Do

2. AI Model Providers

When you send a message to Alfred, your query is forwarded to one of the following third-party AI providers for processing. All providers listed are based in the United States.

2.1 Text & Reasoning Models

2.2 Image Generation Providers

2.3 Other AI Services

3. Your Data and AI Training

3.1 Alfred Does Not Train on Your Data by Default

• We do not use your conversations, uploaded files, or queries to train AI models
• We do not share your data with third-party model providers for training purposes
• Your conversations are stored solely to provide the service (conversation history, context)

3.2 What We Do Use Data For

• Intent classification: We embed your query (using Google Gemini embeddings) to determine which AI model and capability to route to. This embedding is transient and not stored beyond the request.
• Quality review: We may periodically review anonymized queries to evaluate whether Alfred responded well. No personally identifiable information is used in this process.
• Abuse prevention: Request patterns are analyzed to detect and prevent API abuse and trial fraud.

3.3 Per-Provider Training Policies

Training practices differ by provider and by API tier. The following reflects our current understanding based on each provider's published API terms:

We review and update this table when providers update their terms. Please check each provider's current policy directly for the latest information.

4. AI Limitations and Accuracy

4.1 AI Can Make Mistakes

• AI models can produce incorrect, outdated, or misleading information — this is known as "hallucination"
• Always verify important information from authoritative sources
• Alfred is not a substitute for professional legal, medical, financial, or technical advice

4.2 Content Filters

• Alfred applies safety filters from underlying model providers
• Requests that violate our Terms of Service will be refused
• Generated images are subject to content policies from the image provider

4.3 Knowledge Cutoffs

Most AI models have a training data cutoff date. For current events, Alfred uses web search to supplement model knowledge where available.

5. Your Controls

• Delete conversations: Remove any conversation from your history at any time
• Conversation memory: Alfred stores facts inferred from your conversations (e.g. "User is VP of Engineering") to personalise responses. You can disable memory entirely via Settings → Preferences, or delete individual entries via Settings → Memory Management
• Export your data: Download all your conversations via Settings → Account → Export Data
• Delete your account: Permanently removes all your data within 30 days

6. Responsible Use

We are committed to responsible AI deployment:

• We do not enable generation of CSAM or content that sexualizes minors under any circumstances
• We do not enable generation of content designed to facilitate real-world violence
• We monitor for abuse patterns and cooperate with law enforcement when legally required
• We apply each provider's content safety policies in addition to our own

7. Contact

Questions about this AI Usage Policy:

• Email: support@snrblabs.com

Please include the nature of your request (privacy question, safety concern, data request) and we will respond appropriately.

Last Updated: February 25, 2026
Version: 1.1