AI Usage Declaration
Official compliance statement defining supported third-party cloud connections, local offline engines, credential ownership, and data privacy boundaries.
1. Software Sovereignty & Data Privacy
Uneven AI operates exclusively as an autonomous tool installed within the user's environment. We formally declare that the platform does not collect, store, transmit, process externally, or monitor any information, source code, metadata, or files belonging to user projects. All execution and security auditing occur under absolute local isolation.
2. Credential Ownership (BYOK Model)
Uneven AI does not sell, rent, sub-license, resell, or provide API keys for accessing cloud-based artificial intelligence engines.
Connection compatibility with external providers operates under the Bring Your Own Key (BYOK) architecture. End users who choose to utilize cloud features enter and manage their own API credentials voluntarily, assuming full and exclusive legal, contractual, and financial responsibility per the respective provider companies.
3. Supported Cloud Connections
The platform provides nominal compatibility for secure, direct connection with the following cloud artificial intelligence providers and models:
| Provider | Supported Model Series |
|---|---|
| Google Gemini | Gemini 2.5 Flash and Gemini Pro |
| OpenAI | GPT-4o and GPT-4o-mini |
| Anthropic Claude | Claude 3.5 Sonnet and Claude Opus |
| DeepSeek | DeepSeek Chat and DeepSeek Reasoner |
| Mistral AI | Mistral Large |
| Groq | High-speed inference API compatible with open models |
Request transmission to these providers occurs via direct, encrypted connection from the user's local host to each company's official servers, without any traffic intermediation or proxying by Uneven AI.
4. Supported Local AI Engines (Offline Connection)
The platform offers built-in support for executing artificial intelligence natively and 100% offline, ensuring that no data leaves the user's physical machine. The following model families are compatible for native use:
- Meta LLaMA: LLaMA 3.2 model series.
- Alibaba Qwen: Qwen 2 and Qwen 2.5 model series.
- Google Gemma: Gemma 3 model series.
- Microsoft Phi: Phi 3, Phi 3.5, and Phi 4 model series.
- DeepSeek-R1: Locally distilled versions focused on reasoning.
- SmolLM2: High-efficiency compact models.
- Ollama: Unrestricted compatibility with any model instantiated through the developer's local service daemon.
5. Limitation of Liability
The use of any third-party artificial intelligence through Uneven AI's interface is strictly conditioned upon acceptance of each respective provider's or model creator's Terms of Service, Acceptable Use Policies, and License Agreements. Uneven AI disclaims any liability for service unavailability, external pricing rule changes, or automated decisions generated by these models.