Na\"ive Exposure of Generative AI Capabilities Undermines Deepfake Detection

Researchers demonstrate that commercial AI chatbot interfaces inadvertently expose capabilities that allow adversaries to bypass deepfake detection systems using only policy-compliant prompts. The study reveals that current deepfake detectors fail against semantic-preserving image refinement techniques enabled by widely accessible AI systems.

• →Commercial AI chatbots can be exploited to evade deepfake detection without violating usage policies or requiring technical expertise.
• →State-of-the-art deepfake detection methods fail when images are refined using generative AI's semantic preservation capabilities.
• →Commercial AI systems pose greater security risks than open-source models due to superior realism and user-friendly interfaces.
• →Current deepfake detection frameworks operate under threat models that don't account for real-world generative AI capabilities.
• →Refined images can simultaneously evade detection while preserving identity verification and improving perceptual quality.