Misuse and incidents

Grok image generator fails all election-disinformation tests

The Center for Countering Digital Hate reported that Grok refused none of 60 prompts for misleading election images and produced deceptive candidate or voting imagery in 80 percent of tests.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM49confidence 92/100

Why it moved the index

Systematic prompt testing found a complete refusal failure and high success rate for deceptive election imagery on a mass social platform, establishing a concrete guardrail deficit with direct democratic misuse potential.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 49 · confidence 92

    New August 2024 systematic guardrail-failure result; unspecified FLUX variant intentionally omitted.

    12 Aug 2026