SUBSCRIBE

At Black Hat, Google warned of AI-powered attackers. Google's own tracker counts the breakthroughs: zero.

The 'new entry path' on stage is stolen session cookies, a technique older than the models. The genuinely new thing is one AI-built zero-day, and Google caught it before anyone got hit.

01THE CLAIM
"Threat groups are using AI models to develop new exploits and gain entry into corporate networks through new techniques, dragging defenders into an AI-based arms race" [SOURCE ↗]

THE MOVE: ZERO UNDERNEATH, the headline number has nothing behind it

TRUE, BUT5 SOURCES · LIVE 2026-09-05
GOOGLE THREAT INTELLIGENCE GROUP TRACK RECORD1 CLAIM · 40/100 BS RATE →
0BREAKTHROUGH CAPABILITIES IN GTIG'S OWN TRACKER
1AI-BUILT ZERO-DAYS OBSERVED IN THE WILD, EVER
MAY 11WHEN GOOGLE LOGGED THAT ONE
02THE CHECK

THE CLAIM. at a Black Hat media briefing, Google Threat Intelligence Group and Accenture said threat actors are using frontier and open-weight AI models to develop new exploits and new entry paths into corporate networks.

THE CHECK. Google's own written AI Threat Tracker says it 'has not yet observed' actors achieving breakthrough capabilities that alter the threat landscape. The headline 'new technique' is stealing tokens, cookies and session IDs, which infostealers did long before AI. The one confirmed AI-built zero-day was disclosed May 11 and closed before it was used.

03SAY THIS IN THE MEETING
"'Which of these entry paths did not exist before AI?' If the answer is stolen session cookies, that path is older than the models selling the panic."

On August 12 at Black Hat USA in Las Vegas, Ryan Whelan of Accenture and John Hultquist of Google Threat Intelligence Group gave a media briefing on AI-enabled attackers. The message, as reported by Cybersecurity Dive: threat actors are using frontier and open-weight models to 'develop new exploits,

🔒 THE FULL AUTOPSY · FREE WITH AN ACCOUNT

You just read the free check. Sign in free, a code by email, no passwords, and the rest unlocks: the evidence trail, the steelman and the rebuttal, all 5 sources with quotes and screenshots, and our on-record call.

This story is a stable, citable object. If you can falsify a verdict, tell us. Corrections are loud here.