Science & Tech · Cybersecurity · 1 hr ago
Google says suspected Russian-speaking hackers used fully automated AI attack
Google says hackers used AI to run a cyberattack without human help at any stage, which its researchers believe is a first.
The suspected operator was Russian-speaking, but Google did not name a hacking group or state.
The system, called exploit_agent, compared software code before and after security updates to work out what flaws a patch had fixed.
It then created and ran exploits automatically, and the server also held data stolen from victims and scripts for activities such as reconnaissance and password guessing.
Information on the server suggested the operator targeted French media, entertainment and education websites.
Google says automation could make it easier to find and exploit flaws at scale, and it expects attackers to keep using such tools against organizations worldwide.
Google said suspected Russian-speaking hackers used an AI system to discover vulnerabilities and carry out exploits without human involvement.
The system, called exploit_agent, compared software code before and after security patches, then developed and ran exploits automatically.
Google researchers found data stolen from victim organizations on an exposed staging server.
The server’s data suggested the operator targeted French websites in media, entertainment and education.
Google said the activity marked a significant advance in the use of AI for cyberattacks.
- Who
- A suspected Russian-speaking operator; Google did not name a group or state.
- What
- The operator used an AI framework to find vulnerabilities and execute exploits without human involvement.
- When
- Google reported the findings on Nov. 16, 2026. The operation’s date is not stated.
- Where
- Victim data suggested targeting of French websites in media, entertainment and education.
- Why
- Not stated.
This story does not have two clearly opposing sides.
Google counted 22 exploited zero-days, compared with an average of eight per month in 2025.
Vendors patched two PHP vulnerabilities.
Hundreds of purported exploits for the PHP vulnerabilities appeared, most nonfunctional and likely AI-generated.
Google’s findings were reported in its fourth-quarter 2026 AI threat tracker.
- Framework
- exploit_agent
- Suspected operator
- Russian-speaking; no group or state named
- Suggested target locations
- French websites
- Sectors
- Media, entertainment and education
- Zero-days exploited in August 2026
- 22
3 sources
What people are saying
Given the low barrier to entry, ease of integrating new exploits, and massive scaling potential, we believe it is highly likely that threat actors will continue to use these frameworks in campaigns targeting organizations globally.
We have observed several threat actors separately experimenting with use of these frameworks and harnesses, including financially motivated, as well as suspected espionage threat actors with a PRC nexus.