4AIs runs an AI agent with real device control — accessibility gestures, screen reading, and optional shell access via Shizuku. We take security reports seriously.

Use GitHub's private vulnerability reporting (Security tab → "Report a vulnerability") or email sanjay@4ais.in. Please do not open a public issue for security problems. You'll get a first response within 72 hours.

• API keys never ship with the app. Your provider key (NVIDIA / OpenAI / Anthropic / Gemini) is entered in Settings and stored in Android EncryptedSharedPreferences (Keystore-backed). If secure storage can't be initialised, the app refuses to fall back to plaintext and fails loudly.
• No middleman server. Prompts and screen content go directly from the device to the provider you chose. There is no proxy, no account, no telemetry.
• Local-only gateway. The OpenClaw gateway binds to 127.0.0.1 and requires a per-install random token (~/.openclaw/auth-token, mode 0600). The UI-automation bridge on :3001 requires the same bearer token with a constant-time comparison.
• Human-in-the-loop. Agent actions above "read" risk are gated by an approval dialog; sensitive intents always prompt even when auto-approve is on. Every agent action lands in an on-device audit log.

Reports about bypassing any of the above — token leakage, unauthenticated routes on :3000/:3001, approval bypasses, key exfiltration — are exactly what we want to hear about.