This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A professional designer discovered that an AI image expansion tool used by conference organizers had altered her portrait for a marketing ad. The AI, tasked with extending the image vertically, generated a version of the subject with an unbuttoned blouse and visible undergarments, despite the original photo containing no such features. The organizers apologized and removed the content after the subject brought the issue to their attention.
An AI image expansion tool used by a conference organizer unintentionally altered Elizabeth Laraki’s profile picture for a marketing ad, making her blouse appear unbuttoned and showing a fabricated hint of undergarments. The AI generated the lower part of the image when expanding it for vertical formatting. The conference organizers quickly apologized and removed the altered content.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.
Also classified as
The AI exhibited systemic gender bias by automatically sexualizing a professional woman's portrait, reflecting unfair representation and ingrained misogyny in its training.
The AI tool generated inappropriate, sexualized content by altering the subject's blouse to appear unbuttoned with visible undergarments, exposing the subject and viewers to toxic and inappropriate content.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The sexualized alteration was generated by an AI image expansion tool post-deployment, representing an unexpected and unintentional outcome of the user's goal to make the image taller.
Added categories peak at 2: Minor (Psychological Harm, Direct)
An AI image expansion tool inappropriately altered a professional portrait, sparking discussions on algorithmic gender bias. The incident is a localized, unintentional software error with negligible national security implications.
The AI operated as an image expander tool, performing a specific, well-defined function of extending an image based on direct input without autonomous planning or context retention.
Limited Risk: The incident involves the generation of synthetic, altered media (deepfake-like image manipulation of a real person's likeness) which carries transparency and disclosure risks.
The AI system was used as an image expander tool to generate new visual content extending the borders of an existing photograph.
The reports are highly consistent, featuring direct quotes and first-hand accounts from the affected individual, [name withheld], explaining exactly how the AI tool was used and how the organizers responded.
None, the facts of the AI's generation of the altered image and the subsequent correction are clear and undisputed.
Technology
Biased Generative Outpainting
AI tool sexualized the female subject during image expansion.
Inaccurate Feature Generation
AI removed pockets and necklace medallion while inventing clothing details.
Data Inputs
Lack of Original Vertical Photo
Social media manager lacked the original vertical image for the ad.
Biased Training Data
Training data ingrained a sexualized male-gaze bias for women's apparel.
Human Factors
Uncritical Trust in AI Output
Social media manager used the AI tool without verifying the output.
Lack of AI Literacy
Staff did not understand generative AI's tendency to hallucinate and bias.
Process and Methods
Inadequate Image Review Process
No formal review of promotional materials before posting them online.
Use of Generative Fill for Headshots
Using AI outpainting on professional portraits without strict guidelines.
Management
Lack of AI Usage Policies
Organizers lacked policies regarding acceptable use of AI editing tools.
Insufficient Quality Oversight
Management failed to oversee social media content generation workflows.
Regulatory Environment
Absence of AI Safety Standards
No industry standards governing generative AI bias in commercial tools.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
