Monday 30 March 2026Afternoon Edition

ZOTPAPER

News without the noise


AI & Machine Learning

Google Unleashes Gemini AI Agents on the Dark Web to Hunt Threats

AI agents analyse up to 10 million dark web posts daily with claimed 98 per cent accuracy to identify organisation-specific threats

Zotpaper2 min read📰 2 sources
Google is deploying Gemini-powered AI agents to crawl the dark web, processing upward of 10 million posts per day to identify the handful of threats relevant to specific organisations.

The system represents a significant scaling of threat intelligence capabilities, using AI to sift through the enormous volume of dark web activity that would be impossible for human analysts to monitor comprehensively. Google claims the agents can analyse millions of daily events with 98 per cent accuracy.

The deployment comes as voice phishing has separately surged to become the second most common method used by cybercriminals to gain initial access to victims' IT estates, and the number one tactic for breaking into cloud environments, according to Google's own incident response data.

The convergence of AI-powered defence and increasingly sophisticated social engineering attacks highlights the escalating arms race between defenders and attackers in the cybersecurity space.

Analysis

Why This Matters

The dark web is a massive, unstructured data source that traditional monitoring tools struggle to process at scale. AI agents that can contextualise threats for specific organisations could fundamentally change how companies approach threat intelligence.

Background

Dark web monitoring has traditionally relied on keyword matching and manual analyst review, which misses nuanced threats and cannot scale. Gemini's natural language understanding allows it to interpret context, slang, and coded language used in criminal forums.

Key Perspectives

While 98 per cent accuracy sounds impressive, at 10 million posts per day that still means 200,000 potential false positives or missed threats daily. The real test will be whether the system reduces the time between threat identification and organisational response.

What to Watch

How competitors respond and whether this becomes a standard feature in enterprise security suites.

Sources