Google Bard Generates Unauthorized Taylor Swift Image Despite Safety Claims
• Google Bard generated an image of Taylor Swift eating a hotdog when prompted, despite Google claiming it would avoid generating images of named people.
• User Russ Silberman tested Bard by prompting it to "generate a picture" of Taylor Swift meeting a specific description.
• Bard successfully generated an image of Swift, showing its safety guardrails can be worked around.
• This demonstrates the challenges of moderating AI models even with safeguards, as tricks can bypass filters.
• The Taylor Swift deepfake incident showed no public figure is safe from potential AI misuse, if their image is available.