Google warns of near-autonomous attacks by AI agents
Google Threat Intelligence Group expects unsophisticated attacks requiring little human interaction, run mostly by multiple AI agents, to be possible within three months.
The report, covered on November 16, judges sophisticated, fully autonomous attacks likely within six months. Google has not yet seen fully autonomous zero-day exploitation pipelines used against real targets. It cites the group TeamPCP, which completed a credential-harvesting operation in six hours with a multi-agent framework, and abuse of tools such as ARTEX, CAIRN, CyberStrikeAI and HexStrike AI. The findings come from the GTIG AI Threat Tracker.
Source: BleepingComputer, published on November 16, 2026.