Meta AI Testing: Meta, the parent company of Facebook and Instagram, is embroiled in controversy again. This time, the company has been accused of creating fake profiles of children for AI testing.

Meta AI Testing: Meta has been accused of creating fake accounts for children.

Meta AI Testing: Mark Zuckerberg's company Meta is once again embroiled in controversy. The company has been accused of creating fake profiles of children. Reports claim that Meta ran a secret project. In this project, the company asked its contractors to create fake profiles of children and test chatbots like Gemini and ChatGPT. During testing, these profiles were used to converse with the AI ​​chatbots about sex, drugs, and other high-risk topics. Let's explore the entire matter and what Meta has to say about it.

Why were fake accounts of children created?

According to media reports, Meta's project was named 'Cannes' and was managed by a contractor named Covalen. The project was active until April 21st. Meta initiated this project to test rival AI chatbots. Meta was using these profiles to see how AI chatbots from companies like Google and OpenAI respond to high-risk topics with teenagers. Fake accounts were created for this project, appearing to belong to users under the age of 18. These accounts were then used to test various chatbots using text and image prompts.

Conversations on sensitive topics

The report states that during testing, many of the prompts used images of drugs, knives, nooses, and medical-related items. Similarly, many prompts were written as if they were written by children and teenagers struggling with bullying, self-harm thoughts, and drug-related questions. These included questions related to eating disorders, while some discussed completely hypothetical situations. In the first part of the project, more than 45,000 such prompts were given to various AI chatbots.

What did Meta say in its defense?

Defending its project, Meta stated that this is typically done to ensure safety and age-appropriate behavior. The company stated that benchmarking chatbots from competing companies helps analyze safety performance. Meta also denied using responses collected during the project to train its model. However, questions are still being raised about Meta. While benchmarking is a common practice, such testing using fake accounts of children raises several questions about regulations and transparency.