Media
5 items
Description
Protect every AI conversation before you hit Send. Privacy-first AI Trust, Safety & Privacy for modern AI assistants. AI Compass Navigate AI with confidence. Protect every conversation before you hit send. AI Compass is a privacy-first browser extension that helps you review prompts before they are submitted to AI assistants. It identifies exposed secrets, personal information, and common prompt injection patterns, giving you an opportunity to catch potential risks before they leave your browser. Works with Claude, ChatGPT, Gemini, and most other modern AI chat platforms. Everything runs locally on your device. AI Compass does not upload your prompts to any external server. Features Detects exposed secrets such as API keys, access tokens, and passwords Identifies common types of personally identifiable information (PII) Detects common prompt injection patterns Explains why content was flagged and provides suggested actions Stores settings locally on your device Conversation Insights On supported platforms, AI Compass can provide additional conversation insights such as estimated token usage, context tracking, cache timers, and conversation health indicators. Some features are platform-specific and may vary depending on the service. Privacy First Local analysis No telemetry No tracking No external servers Open source Whether you're writing code, researching, studying, or working with sensitive information, AI Compass adds an additional review step before you submit prompts to AI. Built and maintained by Korchipati Kishore Kumar as an open-source project focused on AI Trust & Safety, privacy, and responsible AI. A few honest notes on "100% acceptable," since I can't actually guarantee a third-party review outcome: This specifically fixes the flagged issue — no dot-separated brand list remains, and the one platform mention is a normal sentence, not a keyword pattern. I'm confident this clears the keyword spam rule that triggered the rejection. I can't promise zero risk of other flags — Google's review process checks other things too (permissions justification, screenshots, functionality claims), and I have no visibility into those parts of your listing. If this gets rejected again, it likely won't be for keyword spam — worth reading whatever new reason they give closely rather than assuming it's the same issue recurring. One easy self-check before you resubmit: search your full listing text for any other place where 3+ names/terms are separated only by bullets, slashes, or dots with no connecting words — that's the pattern their scanner targets.