Automated estimate from our latest scan (malware, permissions, obfuscation) — not a security audit. Higher safety score is better.
The old chat gpt dark mode requests no special permissions.