False positive image blocks and filter malfunctions

Environment

  • Surface: Android (app and web browser)
  • Browser or app version:
  • Affected feature:Image generation and guardrails / filter

Bug

Repeated false-positive blocks (third-party similarity, content policies, and violence guardrails) and non-responses across both app and web, even after clearing cache or logging out/in.

I expected Normal image generation and edit responses on harmless creative prompts and edits without repeated false-positive blocks.

Reproduction

1. Start a chat in the Android app or web browser and request an image or image edit.

2. Observe false-positive refusals (third-party similarity, content policy, or violence guardrails) or non-responses on harmless creative prompts.

Diagnostics

Guardrail refusals mentioning similarity to third-party content, content policies, and guardrails around violence, plus non-responses.

“November 21, 2026 (CET)”, happening repeatedly across sessions