Single Image Unlearning (SIU) enables multimodal large language models to forget specific visual concepts, offering a practical approach to implement the 'right to be forgotten' in AI systems.

• Efficiently removes recognition of specific concepts without extensive retraining
• Preserves overall model performance while eliminating targeted visual knowledge
• Protects against membership inference and jailbreak attacks
• Demonstrates effectiveness through comprehensive security evaluations

This research addresses critical privacy and security concerns by giving organizations a practical method to remove sensitive visual data from AI systems, helping comply with privacy regulations while maintaining model utility.

Single Image Unlearning: Efficient Machine Unlearning in Multimodal Large Language Models