At the Google Cloud Security Summit, Google announced a suite of agentic-era security capabilities spanning the development lifecycle, runtime workloads, and security operations. The centerpiece is CodeMender, an autonomous code-security agent now in preview that finds and fixes vulnerabilities using models like Gemini 4 Argon, while new identity and network controls secure autonomous agents across clouds.
CodeMender brings autonomous vulnerability remediation to the IDE
CodeMender enters preview as an autonomous code-security agent that identifies vulnerabilities and generates tested patches before exploits emerge. It works with multiple models including Gemini 4 Argon, which recently helped Wiz discover and remediate core PHP vulnerabilities affecting 70% of websites by finding the flaw, developing an exploit path, detecting attacks, and validating the fix.
New preview capabilities include secure autonomous remediation that isolates vulnerability verification and exploit simulation inside OS-level sandboxes or isolated VMs with human-in-the-loop checks for local writes and command execution. Flexible scanning modes let teams scan specific folders, entire repositories, or only modified files since the last scan to optimize speed and cost.
Hundreds of customers have used CodeMender since its July preview launch. Thales Group reported potential to reduce vulnerability discovery time by over 95% while keeping expert validation under human oversight. IDC noted that nearly 40% of enterprises add vulnerabilities faster than they can remediate them, and CodeMender pairs exploitability validation with deployable code fixes for continuous verified patching.
Auth Manager and identity-based perimeters secure agents everywhere
Auth Manager reaches general availability, centralizing credential management in a Google-managed vault governed by IAM. It automates OAuth 2.0/PKCE flows and refreshes user-delegated tokens to minimize consent fatigue and reduce credential theft. Agent Identity JWT support, now in preview, issues standard OIDC tokens so agents can authenticate securely across AWS, Azure, on-premises backends, and MCP servers without static long-lived API keys.
VPC Service Controls gains identity-based ingress and egress rules with IAM roles (GA) and Workload Identity Federation for GKE (preview). This enables security teams to enforce granular, identity-based perimeter boundaries across enterprise GKE clusters without complex manual exceptions, reinforcing defenses against data exfiltration in flat-trust architectures.
Model Armor adds a custom topic detector in preview, letting teams define prompt-based rules to identify on-topic and off-topic agent interactions. Tailored for healthcare, insurance, and retail use cases, it keeps agent conversations focused, compliant, and secure. Advanced network threat protection expands with Cloud NGFW advanced malware sandbox (GA) powered by Palo Alto Networks WildFire, project-level firewall endpoints (GA), and Cloud Armor managed rules (preview) backed by Thales Imperva's 1.5 trillion monthly web request scans.
Machine-speed defense and Wiz partnership extend coverage
Google Security Operations now combines continuous workload monitoring feeding telemetry into the Wiz Security Graph, Google Threat Intelligence codified into defensive model runtime logic triggering automated agentic threat hunts, and autonomous agents analyzing complex alerts with correlation queries across multi-vendor telemetry in sub-second timeframes. Mandiant Transformation Services helps modernize operating models to maximize these technologies.
The Wiz-Google Cloud partnership deepens with Wiz AI SAST in preview using agentic reasoning to surface business logic flaws, correlating code weaknesses with infrastructure context on the Wiz Security Graph. Exploitable paths can be remediated at machine speed using Wiz Green Agent and CodeMender. Wiz Defend and Google Security Operations now share a data model with bi-directional syncing, a dedicated Wiz Content Pack, and Wiz Blue Agent investigation analysis.
Scan for Good, a joint Google DeepMind and Wiz initiative, brings autonomous technology with human researchers to under-resourced organizations. In its first month, it used Argon to help over 50 public service organizations and critical infrastructure providers find and remediate critical exposures. Google also continues disrupting adversaries by tracking bad actors targeting Google Cloud, engineering specialized defensive architectures with DeepMind, and feeding frontline intelligence back into model alignment.
What to do next
Teams evaluating these capabilities should start with CodeMender preview for development pipeline security, enable Auth Manager GA for credential governance, and review VPC Service Controls identity-based rules for perimeter enforcement. The Google Cloud Security Summit sessions provide deeper technical walkthroughs.
Source: Helping you build and defend at machine-speed in the agentic era (GCP).
