AI done right: transparent, responsible, and client-focused
Ethical Use of AI
HOW WE USE AND DON'T USE YOUR DATA
Data Usage Requires Opt-In
Ready for some nevers? We never use your data for training or testing AI models unless you explicitly opt in. When you opt in, you choose whether your data is used for testing only, or training and testing.
-
Your data may be used to fine-tune Canopy-specific models, making them more useful for your firm and others like yours.
-
Fine-tuned models and training data are never shared externally.
-
Your data is never used to train general-purpose models built by vendors like OpenAI or Anthropic.
We build our own AI tools and also partner with trusted providers like OpenAI and Anthropic. These third-party tools are used only within secure parameters, and they do not train their general models with your data.
AI GOVERNANCE
How We Keep Your Data Safe
We’ve implemented multiple layers of protection to ensure your data remains secure, even when it’s being used to improve our AI tools.
-
PII Handling: Before any data is used, personally identifiable information (PII) is removed using industry-standard techniques, unless it’s essential to completing a specific task.
-
Data Minimization: We only use the minimum data necessary for the specific function the AI is performing—nothing more.
-
Custom and Shared Models: Most of our AI models are custom-built per client, significantly reducing the risk of cross-client data exposure. For shared models, all inputs and outputs follow strict schemas that are continuously updated based on best practices and research.
Our principles for ethical AI:
Our tools are designed to be explainable, auditable, and always under human control.
Transparency
No AI is perfect. That’s why every output is editable, and you make the final call. We track overall model performance to improve behind the scenes, but we never use your firm’s data unless you’ve opted in.
Fairness
Bias has no place in your firm or in your tech. Before releasing any AI functionality, we assess it for potential bias and discrimination. We refine and test our models to ensure they support inclusive, equitable outcomes across all user interactions
Privacy & Security
Canopy is SOC 2 certified, and our shared models follow rigid schemas and continuously updated security protocols.