Google Cloud Forecast: Agentic AI Will Reshape Cyberattacks by 2027

Google Cloud published its "2027 Security Forecast" on November 10, 2026, predicting that agentic AI will accelerate both cyberattacks and defense by 2027 and change what companies need to protect, according to dailysecu.com.

The forecast proposes that autonomous agents' input data, decision rationale, tools invoked, values passed and authorizing policies be recorded so their actions can be verified later. Heather Adkins, Vice President of Security Engineering at Google, said Zero Trust should be extended to individual transactions and requests, with verifiable identity and least privilege applied per agent.

Google Cloud also outlined a security operations center model in which staff can understand the basis of an automated response within 60 seconds and revert it within 30 seconds.

Supply chain and data theft risks

The company warned that open-source libraries pulled in by AI coding tools, along with tools built on MCP, could become targets for supply-chain attacks. It predicted attackers will shift toward data theft and extortion, with AI model weights, agent functions and prompt libraries as targets, alongside unauthorized use of cloud GPU resources, a practice it calls "LLMJacking."

Charles Carmakal, Chief Technology Officer at Mandiant Consulting, recommended red-team exercises to test safeguards and permission controls before autonomous agents are deployed in real work. Google Cloud's Chief Information Security Officer Chris Betz and Chief Product and Customer Officer Karthik Narain are among the executives associated with the forecast.