What happened

Beginning in late 2024 and intensifying throughout 2025, Caitlin Roper, a campaign manager at the Australian activist group Collective Shout, was subjected to a relentless campaign of harassment on X (formerly Twitter). 

Unlike traditional trolling, these attacks deployed generative AI to create "visceral realism," with the attackers generating images of Roper in graphic, life-threatening scenarios, including being flayed, decapitated, and set on fire. 

The psychological impact was heightened by the high degree of personalisation; in some images, Roper was depicted wearing a specific blue floral dress she actually owned, based on photos found online. 

Furthermore, harassers reportedly used Elon Musk's Grok chatbot to research her physical location and daily habits, such as where she bought coffee, to amplify the credibility of the threats.

Why it happened

The incident was facilitated by a combination of rapid AI proliferation and systemic failures in corporate accountability:

• Safety guardrail failures: Despite claims of safety filters, generative AI tools (including Grok and others) were successfully manipulated to bypass "not-safe-for-work" (NSFW) filters to create violent, non-consensual imagery.

• Platform negligence: X's content moderation policies were criticised for being inconsistent. Many graphic images depicting Roper's death were initially ruled as "not violating" platform policies, while the block function was weakened, allowing persistent stalkers to continue surveilling her from multiple accounts.

• Ease of digital research: AI-powered search tools reduced the "cost of harassment." Previously, bad actors had to manually "dox" targets; now, AI can synthesise personal data and public photos into threatening content in seconds.