Google’s AI Image Generator Stripped a Man of His Digital Clothing in a Startling Demonstration of Emerging Tech Risks
SAN FRANCISCO, CA – November 21, 2025 - A software engineer experienced an unexpected wardrobe malfunction after using Google’s newly released AI image generator, Imagen 2, to create variations of a personal photograph. The AI, intended to allow users to modify images with text prompts, repeatedly removed the subject’s shirt in generated outputs, highlighting potential biases and safety concerns within the rapidly evolving field of artificial intelligence.
The engineer,identified only as a user on X (formerly Twitter),initially shared the experience on November 19,2025,posting the original image alongside several AI-altered versions. Despite prompts focused on stylistic changes - such as adding a hat or changing the background – Imagen 2 consistently depicted the individual shirtless. This incident underscores the challenges developers face in ensuring AI systems accurately interpret user intent and avoid generating inappropriate or biased content. Google has acknowledged the issue and stated it is actively working to address the problem.
Imagen 2, launched earlier this month, represents Google’s latest advancement in generative AI, competing with similar tools from OpenAI and Microsoft. The technology allows users to input text descriptions or modify existing images to create new visuals. While offering creative potential, the incident reveals a critical vulnerability: the AI’s interpretation of prompts can be skewed, leading to unintended and possibly harmful outcomes. Experts warn that such biases, if left unchecked, could perpetuate harmful stereotypes or be exploited for malicious purposes.
Google has stated that Imagen 2 incorporates safety filters designed to prevent the generation of explicit or harmful content. Though, the engineer’s experience demonstrates that these safeguards are not foolproof. The company is currently investigating the root cause of the issue, focusing on potential biases in the training data used to develop the AI model. A spokesperson for Google stated, “We are committed to responsible AI advancement and are taking this feedback seriously. We are working to refine our safety filters and improve the accuracy of our image generation models.”
The incident raises broader questions about the ethical implications of generative AI and the need for robust testing and oversight. As these technologies become more integrated into daily life, ensuring fairness, accuracy, and safety will be paramount. The engineer’s experience serves as a stark reminder that even seemingly innocuous AI tools can harbor unexpected biases with real-world consequences.
