The End of AI "Slop"? How Google is Using LoRA and LLMs to Fight Coordinated Synthetic Spam

Short summary

Google researchers introduce S-CTS, an AI defense system that detects coordinated spam networks by analyzing account behavior and content artifacts rather than individual videos. Using LLMs with LoRA and Automatic Prompt Optimization, the system adapts to new AI spam trends in days, achieving 32–50% faster review times while maintaining 92–95% precision. This network-centric approach offers a scalable template for fighting AI-generated slop.

• •S-CTS shifts moderation from individual content to coordinated account networks using LLMs with LoRA
• •Achieves 32–50% faster validation and review times with 92–95% precision to avoid false positives
• •APO-based prompt optimization allows the system to adapt to new AI spam generators in days rather than months

Generated with AI, which can make mistakes.