Direct answers to how AixSafe works, our security model, and compliance positioning.
How does AixSafe work?
AixSafe acts as a vendor-agnostic proxy that sits between your AI applications (like Copilot or custom agents) and LLM providers. It intercepts requests, scrubs PII, enforces policies, and logs telemetry in real-time. No changes are required to your application code — simply point your AI endpoint to AixSafe.
Is the proxy-first approach secure for banking?
Yes. By being proxy-first, PII and sensitive internal data are redacted at the network layer before ever reaching an external LLM provider. This satisfies strict banking data residency and privacy requirements, including APRA CPS 234 and applicable GDPR obligations.
Do I need to change my application code to use AixSafe?
No. You simply point your AI application's API endpoint to the AixSafe proxy. It is designed to be drop-in compatible with standard LLM provider APIs (OpenAI, Anthropic, Gemini). Integration typically takes under an hour.
What LLM vendors do you support?
We are truly vendor-agnostic. We support OpenAI, Anthropic, Google Gemini, Mistral, and local LLM deployments (Llama, Falcon). Any provider with a REST-based API can be monitored and secured via AixSafe.
Does AixSafe store the prompts and completions?
AixSafe stores hashed telemetry metadata in an encrypted, append-only audit log. The content of prompts and completions is stored only within your organization's own infrastructure — AixSafe does not custody raw PII or model outputs on its own servers.
What compliance standards does AixSafe help with?
AixSafe is designed to support SOC 2 Type II reporting, EU AI Act Article 9 risk management requirements, OWASP AI Top 10 (LLM01–LLM10), and internal governance frameworks. We provide exportable audit evidence packs on request.
Platform compliance positioning
AixSafe provides orchestration, routing, and reporting. Model inference, data storage, and fund custody remain with your chosen providers and cloud infrastructure.
AixSafe acts as an orchestration and reporting platform only
Policy engine operates on data in transit, not long-term storage
AixSafe does not custody customer PII, model weights, or inference outputs
All regulated activities are performed by your existing licensed infrastructure providers
Talk to our team
For detailed compliance documentation, audit evidence packs, or security architecture reviews, contact our engineering team directly.