AI Leaders Join Forces to Safeguard Children in AI Applications

OpenAI, Anthropic and Meta collaborate to support AI safety standards against AI-generated harmful imagery

Ben Wodecki, Jr. Editor

April 30, 2024

2 Min Read
A shadow illuminated onto frosted glass
Getty Images

Some of the biggest names in AI including OpenAI, Google and Meta have pledged to develop safety measures to prevent generative AI systems from being misused against children.

One of the darker uses of generative AI has been predators utilizing the technology to generate abusive imagery. Research from the nonprofit Internet Watchdog Foundation published last October found that this AI-generated imagery was convincing enough to be real.

AI developers scour the web for data but some sources can be unintentionally tainted with such content. The LAION dataset, used by popular AI models like Stable Diffusion, was found to contain thousands of abusive images, according to December 2023 research from Stanford University’s Internet Observatory.

Thorn, an online child safety group, has previously uncovered that open source generative AI systems can be used to create such content.

Thorn has now enlisted the commitment of Stability AI, who co-developed Stable Diffusion, along with other major model developers to address the issue.

The likes of Microsoft, Amazon and Anthropic have committed their support to Thorn’s Safety by Design principles, which encourages models to mitigate related risks at every stage of AI development.

Thorn’s Safety by Design principles are designed to cover the various stages of an AI system’s life cycle, spanning development to deployment.

Related:AI Safety Benchmark for Evaluating Language Model Risks Unveiled

For the development stage, signatories are encouraged to responsibly source datasets, checking corpora for any abuse imagery.

Model developers should also conduct firm stress-testing, rigorously evaluating a system’s generation capabilities throughout the development process.

Thorn also wants firms building generative AI to use detection solutions so images can be traced back to a model. Meta and Google are among the companies supporting Thorn’s efforts to create watermarking systems for their image-generation models.

The nonprofit also encourages generative AI developers to responsibly host AI systems and perform routine red-teaming tests even once they have been deployed.

“The collective commitments by these AI leaders should be a call to action to the rest of the industry,” according to a Thorn announcement.

“We urge all companies developing, deploying, maintaining and using generative AI technologies and products to commit to adopting these Safety by Design principles and demonstrate their dedication to preventing the creation and spread of child sexual abuse material, AI generation child sexual abuse material and other acts of child sexual abuse and exploitation.”

Related:Nightshade Tool Safeguards Images Against Unauthorized AI Generation

ChatGPT maker OpenAI was among the supporters, with the Microsoft-backed company saying it has made “significant efforts to minimize the potential for our models to generate content that harms children.”

Chelsea Carlson, OpenAI’s technical program manager for child safety said OpenAI cares deeply about the safe and responsible uses of its tools, “which is why we’ve built strong guardrails and safety measures into ChatGPT and DALL-E.”

“We are committed to working alongside Thorn,” Carlson said. “All Tech is Human and the broader tech community to uphold the Safety by Design principles and continue our work in mitigating potential harms to children.”

Read more about:

ChatGPT / Generative AI

About the Author(s)

Ben Wodecki

Jr. Editor

Ben Wodecki is the Jr. Editor of AI Business, covering a wide range of AI content. Ben joined the team in March 2021 as assistant editor and was promoted to Jr. Editor. He has written for The New Statesman, Intellectual Property Magazine, and The Telegraph India, among others. He holds an MSc in Digital Journalism from Middlesex University.

Keep up with the ever-evolving AI landscape
Unlock exclusive AI content by subscribing to our newsletter!!

You May Also Like