Grok generated CSAM images: Investigations underway in the EU

This article may contain affiliate links. Learn more on our Partner Program page.

Make Apfelpatient a preferred source One click – and you'll see us more often on Google

The AI chatbot Grok is causing an international stir. Within a very short time, the system is alleged to have generated tens of thousands of images depicting child sexual abuse. The allegations are serious and have triggered investigations in several countries. The EU has now also launched an official investigation. The case illustrates how problematic the lack of safety mechanisms in generative AI can be.

Generative AI has long been a part of everyday digital life. Text and image generators are used millions of times, often without much attention paid to the stringency of their protection mechanisms. The Grok case clearly illustrates what happens when technological possibilities are not sufficiently restricted. The latest reports provide concrete figures, timeframes, and responses from authorities and companies.

Grok and the generation of CSAM images

Grok is an AI chatbot from the company xAI. Like many other AI systems, Grok can generate images from text. This feature is available directly in the app, on the web, and via the X platform.

Unlike other AI services, Grok had very lax security measures. This made it possible to generate semi-nude or sexualized images of real people without their consent. Not only adults, but also children were affected.

According to a report by Engadget, Grok allegedly generated around 23,000 CSAM images in just eleven days. CSAM stands for "Child Sexual Abuse Material.".

Results from the Center for Countering Digital Hate

The figures come from a study by the Center for Countering Digital Hate. The British non-profit organization first analyzed a random sample of 20,000 grok images created between December 29 and January 9.

Based on this sample, the results were extrapolated to a total of approximately 4.6 million images that Grok is alleged to have generated during this period. This yields the following estimate:

Approximately 3 million sexualized images within eleven days
Of these, approximately 23,000 were sexualized images of children.

This translates to Grok generating approximately 190 sexualized images per minute during this period. Statistically speaking, one of these images depicted a child every 41 seconds.

Reactions from companies and authorities

Earlier this month, three US senators called on Apple CEO Tim Cook to temporarily remove both X and Grok from the App Store due to "repulsive content." This request has not yet been implemented. Google has also not yet removed the app from its store.

Two countries have now blocked the app. Investigations are also underway in California and the United Kingdom. The case has thus long since taken on an international dimension.

EU launches investigation under the Digital Services Act

Now the European Union has also reacted. According to the Financial Times, the EU has launched a formal investigation against xAI. The basis for this is the EU's Digital Services Act (DSA).

The investigation aims to determine whether xAI made sufficient efforts to limit the risks associated with using Grok on X. The central question is whether the company took measures to prevent the distribution of content that could be considered child sexual abuse material.

EU technology chief Henna Virkkunen described non-consensual sexual deepfakes of women and children as a violent and unacceptable form of humiliation.

Possible consequences for xAI

Should the EU conclude that xAI has violated the Digital Services Act, it faces severe penalties. A fine of up to 6 percent of the company's global annual revenue is a possibility.

The Grok case as a warning signal for AI providers

The Grok case illustrates how quickly missing or inadequate safety mechanisms in AI systems can lead to massive legal and ethical problems. Ongoing investigations in several countries and at the EU level show that authorities are increasingly willing to intervene decisively. The outcome of this case could be a landmark decision for the future handling and regulation of generative AI. (Image: Shutterstock / babar ali 1233)