Google deploys SAFE AI system to combat AI spam networks

Image credit: Search Engine Journal

Google deployed its new artificial intelligence spam detection system, Scaled Abuse Forensics Examiner (SAFE), globally Tuesday to identify AI-generated content and coordinated abuse networks that mimic human review.

The advanced system marks Google‘s second identified effort in 2026 to combat AI-generated spam, following the introduction of the Scalable Cluster Termination System (S-CTS).

SAFE is designed to detect content that violates policy guidelines, utilizing a few-shot-trained large language model (LLM) to catch material that traditional classifiers might miss, Google reported.

The system operates on three core technical principles: detecting inorganic behavior, automating forensic analysis through multi-agent systems, and employing transformer-based content understanding for policy enforcement.

It incorporates four specialized AI agents to analyze various aspects of potential spam. These agents include a Root Agent, a Content Understanding Agent, a Behavior Understanding Agent, and a Channel Cluster Understanding Agent.

These specialized agents work together to analyze content, user behavior, underlying infrastructure, and the relationships within suspected spam networks, according to Google.

The company stated that SAFE aims to move beyond simple content detection to identify complex, coordinated AI-generated abuse networks that pose a growing challenge to online platforms.

Google has been increasingly focused on developing sophisticated AI tools to counter the rise of AI-generated misinformation and spam, which can quickly evolve to evade existing detection methods.

Source: Search Engine Journal