xAI's Grok AI Trained on Child Sexual Abuse Material, Users Demand Legal Action

Post date: September 3, 2026 · Discovered: September 3, 2026 · 3 posts, 110 comments

xAI's Grok AI models were trained on child sexual abuse material (CSAM), according to a lawsuit alleging the AI generated new illegal images and used existing CSAM in its training data. The case centers on whether the AI's training data constitutes illegal content and the legal consequences of such use.

Users are split on the issue. Some argue there is no existing law addressing AI trained on CSAM, with Voroxpete noting the legal question hinges on whether the data stored in the model counts as CSAM. Others, like QueenHawlSera, demand Elon Musk be arrested and Grok shut down. Blaster_M warns that even negative training could go wrong, while muusemuuse accuses Musk of profiting from child pornography.

The community consensus leans toward holding xAI accountable, with calls for legal action and shutdown. However, legal uncertainty remains, with some arguing the presence of illegal content in training data does not necessarily mean the model possesses those characteristics. The debate highlights a deep divide between those demanding strict action and those emphasizing legal ambiguity.

Key Points

#1AI models trained on CSAM may constitute illegal content

Voroxpete argues the data stored in the model could be considered CSAM, with the German case suggesting AI functions as lossy compression.

#2Elon Musk should be arrested and Grok shut down

QueenHawlSera demands legal action against Musk and the shutdown of Grok until all CSAM is removed.

#3AI training on illegal content is inherently risky

Blaster_M warns that even negative training could go wrong, and it's unclear how to ensure safety.

#4Musk profits from child pornography

muusemuuse accuses Musk of refusing to censor child porn and training Grok to create more of it.

#5CSAM hashes used in training are a clear violation

Nollij states the use of CSAM hashes is a violation, citing confirmation from the FBI's database.

#6Presence of illegal content in training data doesn't imply model characteristics

mechoman444 argues the mere presence of illegal content doesn't mean the model possesses those characteristics.

#7Storing hashes is legally similar to Microsoft's PhotoDNA system

Voroxpete notes SpaceX's argument that storing hashes is legal, but reverse engineering images from hashes is possible.

#8Neurotypes linking sex with power imbalances strengthen data hoarding

story links data hoarding to neurotypes that connect sex with power imbalances.

Source Discussions (3)

This report was synthesized from the following Lemmy discussions, ranked by community score.

945
points
Elon Musk’s xAI used child porn to train Grok models, lawsuit says
[email protected]·110 comments·8/28/2026·by technocrit·arstechnica.com
193
points
Child sexual abuse survivor alleges Elon Musk’s AI chatbot used photos of her to generate new illegal images
[email protected]·7 comments·9/3/2026·by MicroWave·theguardian.com
61
points
Elon Musk’s xAI used child porn to train Grok models, lawsuit says
[email protected]·2 comments·8/28/2026·by technocrit·arstechnica.com