Child abuse images removed from AI image-generator training source, researchers say

Advertisement

Advertise with us

Artificial intelligence researchers said Friday they have deleted more than 2,000 web links to suspected child sexual abuse imagery from a database used to train popular AI image-generator tools.

Read this article for free:


or

Already have an account? Log in here »

To continue reading, please subscribe:

Subscribe and receive a limited-edition Free Press branded hat or tote.

Digital Subscription

One year of digital access for only $205*

  • Enjoy unlimited reading on winnipegfreepress.com
  • Read the E-Edition, our digital replica newspaper
  • Access News Break, our award-winning app
  • Play interactive puzzles

*First annual payment billed as $205.00 + GST for one year. This annual subscription will automatically renew at $233.00 + GST every 52 weeks (10% off the regular annual price of $259.35). Offer available to new and qualified returning subscribers only. Cancel any time.

To continue reading, please subscribe:

Add Free Press access to your Brandon Sun subscription for only an additional

$1 for the first 4 weeks*

  • Enjoy unlimited reading on winnipegfreepress.com
  • Read the E-Edition, our digital replica newspaper
  • Access News Break, our award-winning app
  • Play interactive puzzles
Start now

*Your next Brandon Sun subscription payment will increase by $1.00 and you will be charged $17.95 plus GST for four weeks. After four weeks, your payment will increase to $24.95 plus GST every four weeks.

Hey there, time traveller!
This article was published 30/08/2024 (718 days ago), so information in it may no longer be current.

Artificial intelligence researchers said Friday they have deleted more than 2,000 web links to suspected child sexual abuse imagery from a database used to train popular AI image-generator tools.

The LAION research database is a huge index of online images and captions that’s been a source for leading AI image-makers such as Stable Diffusion and Midjourney.

But a report last year by the Stanford Internet Observatory found it contained links to sexually explicit images of children, contributing to the ease with which some AI tools have been able to produce photorealistic deepfakes that depict children.

FILE - Students walk on the Stanford University campus on March 14, 2019, in Stanford, Calif. (AP Photo/Ben Margot, File)
FILE - Students walk on the Stanford University campus on March 14, 2019, in Stanford, Calif. (AP Photo/Ben Margot, File)

That December report led LAION, which stands for the nonprofit Large-scale Artificial Intelligence Open Network, to immediately remove its dataset. Eight months later, LAION said in a blog post that it worked with the Stanford University watchdog group and anti-abuse organizations in Canada and the United Kingdom to fix the problem and release a cleaned-up database for future AI research.

Stanford researcher David Thiel, author of the December report, commended LAION for significant improvements but said the next step is to withdraw from distribution the “tainted models” that are still able to produce child abuse imagery.

One of the LAION-based tools that Stanford identified as the “most popular model for generating explicit imagery” — an older and lightly filtered version of Stable Diffusion — remained easily accessible until Thursday, when the New York-based company Runway ML removed it from the AI model repository Hugging Face. Runway said in a statement Friday it was a “planned deprecation of research models and code that have not been actively maintained.”

The cleaned-up version of the LAION database comes as governments around the world are taking a closer look at how some tech tools are being used to make or distribute illegal images of children.

San Francisco’s city attorney earlier this month filed a lawsuit seeking to shut down a group of websites that enable the creation of AI-generated nudes of women and girls. The alleged distribution of child sexual abuse images on the messaging app Telegram is part of what led French authorities to bring charges on Wednesday against the platform’s founder and CEO, Pavel Durov.

Durov’s arrest “signals a really big change in the whole tech industry that the founders of these platforms can be held personally responsible,” said David Evan Harris, a researcher at the University of California, Berkeley who recently reached out to Runway asking about why the problematic AI image-generator was still publicly accessible. It was taken down days later.

Report Error Submit a Tip

More Stories

Impaired driving victim honoured

Morgan Modjeski 5 minute read Preview

Impaired driving victim honoured

Morgan Modjeski 5 minute read Saturday, Aug. 15, 2026

Ten years after he was killed in a collision involving an impaired driver in 2016, Robbie Anderson’s name has been added to MADD Canada’s Manitoba Memorial honouring those killed in drunk and drug-impaired-related crashes.

His older brother Chris Anderson, was one of two dozen people who gathered at the Glen Eden Funeral Home and Cemetery in West St. Paul Saturday to remember their loved ones, 86 victims whose names are etched into the granite monument near the heart of the graveyard.

“It’s heavy,” said Anderson when asked what it was like to be part of the ceremony and see his brother’s name added. “It’s been something I’ve been thinking about for a while.”

Grateful for MADD’s work and its efforts to honour Robbie, who was 27 when he died, Anderson said his brother, represented the “absolute best of us.”

Read
Saturday, Aug. 15, 2026

Manitoba honey industry buzzing as U.S. tariff clock ticks

Gabrielle Piché 5 minute read Preview

Manitoba honey industry buzzing as U.S. tariff clock ticks

Gabrielle Piché 5 minute read Yesterday at 6:11 PM CDT

Canadian product is being rushed to the United States ahead of a threatened 50 per cent tariff on the industry.

Read
Yesterday at 6:11 PM CDT

Bombers need all three phases to step up in second half: O’Shea

Taylor Allen 6 minute read Preview

Bombers need all three phases to step up in second half: O’Shea

Taylor Allen 6 minute read Yesterday at 6:44 PM CDT

Highlight-reel catches are a daily occurrence for Ontaria Wilson at practice.

The Winnipeg Blue Bombers receiver also makes jaw-dropping grabs on game days — just go back to last October in Edmonton when Wilson made arguably the most impressive play of the 2025 campaign.

For those that need a reminder, quarterback Zach Collaros underthrew a pass to the end zone that ricocheted off a pair of defenders. Wilson reached out with his right hand and batted the ball up twice while falling to the turf before securing it for an incredible 35-yard score.

That was likely the hardest touchdown of his career, and he should’ve had his easiest in last Friday’s 33-21 win over the Ottawa Redblacks at home — but he dropped it.

Read
Yesterday at 6:44 PM CDT

Hometown hero hits CEBL championship clinching shot

Taylor Allen 5 minute read Preview

Hometown hero hits CEBL championship clinching shot

Taylor Allen 5 minute read Sunday, Aug. 16, 2026

Chapter 1 of Winnipeg Sea Bears basketball was capped by hometown hero Chad Posthumus.

It was May 27, 2023, and the late, great centre out of River East Collegiate had a put-back basket off a Teddy Allen miss to give the Sea Bears a 90-85 victory over the Vancouver Bandits in Winnipeg’s inaugural game, in front of what was then a record crowd of 7,303.

So, on Saturday, it was rather poetic that Simon Hildebrandt — another Winnipegger and a player Posthumus used to mentor — was the one to put an exclamation mark on the franchise’s latest and greatest chapter.

Hildebrandt drilled a wide-open three in Target Score Time to crown the Sea Bears as the 2026 CEBL champions.

Read
Sunday, Aug. 16, 2026

City crews cleaning up drug paraphernalia, weapons having some impact on inner-city life

Malak Abas 7 minute read Preview

City crews cleaning up drug paraphernalia, weapons having some impact on inner-city life

Malak Abas 7 minute read Yesterday at 5:49 PM CDT

A three-person crew, employed by the city as part of its recently expanded proactive cleanup program, allowed the Free Press to follow along last week while they worked to make the Freight House Recreation Centre's property safe to play in.

Read
Yesterday at 5:49 PM CDT

Child abuse images removed from AI image-generator training source, researchers say

Matt O'brien, The Associated Press 3 minute read Preview

Child abuse images removed from AI image-generator training source, researchers say

Matt O'brien, The Associated Press 3 minute read Friday, Aug. 30, 2024

Artificial intelligence researchers said Friday they have deleted more than 2,000 web links to suspected child sexual abuse imagery from a database used to train popular AI image-generator tools.

The LAION research database is a huge index of online images and captions that’s been a source for leading AI image-makers such as Stable Diffusion and Midjourney.

But a report last year by the Stanford Internet Observatory found it contained links to sexually explicit images of children, contributing to the ease with which some AI tools have been able to produce photorealistic deepfakes that depict children.

That December report led LAION, which stands for the nonprofit Large-scale Artificial Intelligence Open Network, to immediately remove its dataset. Eight months later, LAION said in a blog post that it worked with the Stanford University watchdog group and anti-abuse organizations in Canada and the United Kingdom to fix the problem and release a cleaned-up database for future AI research.

Read
Friday, Aug. 30, 2024