ApiPass offers access to models such as GPT and Claude at very competitive prices. For example, our recently added GPT-6 Astra model is available to enterprise customers at 10% of the standard price: https://apipass.dev/model/gpt-6/gpt-6-astra
This pricing advantage comes from the architecture behind our low-cost API, which uses a pool of accounts.
That said, ApiPass is still an API relay. Although we provide trust documentation (https://apipass.dev/trust), it can be difficult to completely address concerns about data security when using a relay. Some providers do use extremely low prices to attract users and then monetize the data they collect from them.
A well-known security researcher recently demonstrated this type of risk: when you use an API relay, credentials and other sensitive information may be exposed.
Image source: https://x.com/shoucccc/status/2098169782541631871
ApiPass does not store user data; we retain only usage information such as token counts. However, trust is inherently difficult to prove conclusively. The only reliable approach is to treat every API relay as untrusted.
So, if API relays should not be trusted by default, how can you use ApiPass safely?
The answer is to build your own intermediary proxy layer and replace all sensitive information with appropriate placeholders.
Several open-source projects already offer this type of solution. These tools were originally designed to prevent companies such as OpenAI and Anthropic from receiving sensitive data, but the same approach is equally applicable to API relays.
At their core, these tools run a local intermediary proxy. When you use CLI tools such as Codex or Claude Code, the proxy scans requests for sensitive information and replaces it with placeholders. The model therefore never receives the sensitive data, while still being able to analyze the request logically. Once the model returns a result, the proxy replaces the placeholders with the original sensitive values, protecting that information throughout the process.
Here are several open-source tools that can meet this need:
- https://github.com/GuthL/KeyClaw
- https://github.com/outgate-ai/og-local
- https://github.com/ahmedxuhri/ai-noleak
- https://github.com/paroque28/claude-code-redact
With tools like these, you do not need to worry about an API relay—or even the official services of OpenAI or Microsoft—accessing your sensitive information.
API relays here do not refer only to ApiPass: you can apply this approach with any relay provider. Naturally, we sincerely hope you choose ApiPass. With its cost, reliability, and security, ApiPass is a strong option.