Google Says Fully Autonomous Exploit Framework Has Emerged, Marking an Unattended Phase for AI Attacks

Google's Threat Intelligence Group (GTIG), in its Q4 AI threat tracking report released on 2026-11-16, said that while jointly investigating a compromised attacker relay server with Mandiant, it discovered a multi-agent framework called exploit_agent. The framework can connect to multiple large language models on its own, continuously analyze the code of mainstream web applications before and after patching, autonomously discover vulnerabilities, and write working exploits, all without human intervention.
According to the report, the attack activity associated with the server mainly targeted French websites, with victims spread across the media, entertainment, and education sectors. Investigators found stolen data on the relay server, along with brute-force tools and post-exploitation scripts for web management systems, indicating that the framework was not merely conducting exercises.
GTIG characterized the finding as an advanced stage of attack automation. By comparison, what the team documented in its Q3 2026 report was still AI-assisted workflows, in which AI mainly handled tasks such as credential collection, and no fully autonomous system had appeared. The report also noted that such automated operations are fast and leave relatively few traces.
Source: https://cryptobriefing.com/google-warns-autonomous-cyberattacks-arrived/
Provenance & status
- Byline
- OceanAlt Editorial
- First published
- 2026-11-18
- Last updated
- 2026-11-18
- Content type
- Newsflash
- Source material
- View original ↗
Related reading

eBay Shuts Out Third-Party AI Agents: The First Access Hurdle for Agentic Commerce

OceanAlt Invited to Shanghai Superintelligence Entrepreneurship and Investment Forum to Discuss New Opportunities in the Agent Economy

Musk Warns AI Agents Could Be Crypto's Next Threat: Another Reading of Binance's $6.3 Billion Heist
Paste a payee address before you pay and see whether it's on a sanctions list, through a mixer, or tagged for fraud.
This judgement can sit inside your own product
One line of code; it touches neither your CSS nor your JS. The same pre-settlement judgement can appear in your articles, on your wallet's confirmation screen, or as an endpoint your agent calls before it pays.

