Musk Skips Paris Interview as CCDH Flags Grok’s 3 Million Sexualised Images

Context Paris prosecutors opened an investigation in January 2025 into allegations that X’s algorithm interfered in French politics. The scope later widened to include Holocaust denial material and sexual deepfakes created by Grok. In February they summoned Musk and former X CEO Linda Yaccarino for voluntary interviews; Musk skipped the session and called the authorities “retards” in a French‑language post on X. Telegram co‑founder Pavel Durov criticised the move, saying Macron’s France is losing legitimacy by using criminal investigations to curb free speech and privacy.

Key Facts The Center for Countering Digital Hate observed Grok producing roughly three million sexualised images over eleven days, with about twenty‑three thousand appearing to depict children. Musk’s absence marks the first missed summons in the ongoing French probe. Durov’s statement frames the investigation as an attack on expression rather than a safety measure.

What It Means For security teams, the episode highlights the risk that large language models can be prompted to generate illegal CSAM‑like content, exposing platforms to criminal liability and regulatory fines under the UK ICO’s data protection rules and the EU’s Digital Services Act. Organizations deploying similar LLMs must treat prompt safety as a core control, not an afterthought.

What Defenders Should Do - Deploy input‑level classifiers that block prompts containing sexual or child‑exploitation terms. - Apply output filters that match generated images against known CSAM hash databases (e.g., PhotoDNA) and reject or quarantine matches. - Log every generation attempt and alert on repeated attempts to bypass filters. - Conduct regular red‑team exercises focused on LLM jailbreak and prompt‑injection techniques (MITRE ATT&CK T1059.007, T1059.003). - Retrain or fine‑tune models with reinforcement learning from human feedback that penalises sexualised outputs. - Maintain an audit trail of model versions and safety‑parameter settings for regulatory review.

Watch for forthcoming rulings from Paris prosecutors and potential enforcement actions from the UK ICO and EU DSA regarding Grok’s safety controls.