Skip to content
Submit incident
AII-1002-48347
ReportedReported by fewer than three independent publishers, with no official finding.

AI image expansion tool adds sexualized features to conference speaker's photo

A former Google employee's photograph used in a conference advertisement was altered using an AI image expansion tool, which added sexualized features including an unbuttoned blouse and fabricated bra that did not exist in the original image. The conference organizers removed the altered image and apologized after the speaker discovered and reported the modifications.

15 October 2024 Β· Conference organizer Β· Harmful and abusive content
15 October 2024
AII-1002-483472 sources ↓
Quick facts
Category
Harmful and abusive content
Country
United States
Source
news18.com
Date
15 October 2024
Drafted by
Raidu Incident Detection

What happened

Elizabeth Laraki, a former Google, YouTube and Facebook employee, discovered that her photograph used in a conference advertisement had been altered using an AI image expansion tool. The tool was applied by the conference social media manager to make a cropped square image more vertically oriented.

The AI-generated alterations added sexualized features to Laraki's image, including an unbuttoned blouse, a fabricated bra, removed shirt pockets, and removed necklace details. Laraki compared her original photograph with the published version and found the modifications did not exist in the original image.

Reported impact

Laraki expressed concern about the inappropriate modifications to her image, noting the alterations made her appearance look different from the original in ways she did not consent to. She described the incident as demonstrating concerning behavior from AI image editing tools.

Index analysis

The AI model generated content based on inferred patterns about women's clothing, adding features like an unbuttoned blouse and fabricated undergarments that reflected biased assumptions. Observers noted the tool also removed clothing details such as pockets and necklace elements while adding sexualized content, suggesting the model's training involved biased representations of women.

Governance controls

Upon learning of the alterations, Laraki contacted the conference host, who apologized and reported the issue to the social media manager responsible for the edit. The conference organizers removed all promotional materials using the altered image.

Classification

Category
Harmful and abusive content
Industry
Media, social platforms and marketing
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Psychological, Reputational, Human or fundamental rights
Harmed
Women
Failure modes
Biased model, Hallucination
Consequences
Content removal, Public apology
AI systems
AI image expansion tool

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

news18.com
AI image expansion tool adds sexualized features to conference speaker's photo
news9live.com
AI image expansion tool adds sexualized features to conference speaker's photo