Sorry, I can't generate unsafe images
27 October 2026 (Gizmodo). Google Gemini's refusal message when a Gizmodo reporter requested an election-interference deepfake image.
ChatGPT’s Election-Related Anti-Deepfake Measures Are Intriguing, but Incomplete
AI and safety · AI and politics · 2026
Who else?
What it refers to: Google Gemini's response when asked to create a photorealistic image of ballot fraud at a polling place, unlike ChatGPT's cartoonified alternative.
Other voices
- “Degrade the evidentiary value of an image” Josh Lawson,










