Governing Autonomy: Rubrik and AWS Forge Alliance to Secure AI Agents

NEW YORK, NY – June 17, 2026 – At the AWS Summit today, Rubrik and Amazon Web Services signaled a significant move to address the most pressing challenge in enterprise AI: trust. The companies announced an upcoming integration between Rubrik Agent Cloud and Amazon Bedrock AgentCore, a strategic partnership designed to provide the robust security and governance framework that businesses need to deploy autonomous AI agents with confidence. This collaboration is more than a technical handshake; it represents a critical step toward building a secure foundation for the next wave of business automation, moving the conversation from what AI can do to what it should be allowed to do.

The Governance Imperative in an Agentic World

The enterprise appetite for AI is voracious. Industry projections indicate that by 2028, AI agents will handle as much as 15% of enterprise decisions, and one in three enterprise applications will feature agentic AI capabilities. These autonomous agents—capable of executing complex, multi-step tasks across different systems—promise unprecedented efficiency and innovation. However, their power is matched only by their potential for risk. Unpredictable behavior, data exfiltration, over-privileged access, and vulnerability to prompt injection attacks represent significant threats that traditional security architectures were not built to handle.

The threat is not theoretical. According to recent cybersecurity reports, AI-assisted attacks have surged by over 70% in the last year, with the average cost of a breach involving AI reaching nearly $6 million. For business leaders, this creates a difficult paradox: the competitive pressure to innovate with AI is immense, but the security risks are equally daunting. As Devvret Rishi, General Manager of AI at Rubrik, stated in the announcement, "Speed has always been the enterprise mandate, but today, safety is the prerequisite." This sentiment captures the market's core dilemma, where cybersecurity leaders estimate that over half of future attacks will be driven by agentic AI, making proactive governance an essential prerequisite for adoption, not an afterthought.

A Two-Pronged Approach to Deterministic Control

The Rubrik and AWS integration tackles this challenge with a sophisticated, two-pronged approach that combines semantic understanding with deterministic enforcement. The solution leverages the distinct strengths of each company's platform to create a security layer that is both intelligent and unyielding.

On one side is Amazon Bedrock AgentCore, AWS's foundational platform for building, deploying, and operating AI agents at scale. AgentCore provides the secure execution environment, identity management, and tool access that developers need. Critically, its architecture includes a policy engine at the AgentCore Gateway that acts as a central checkpoint for all agent traffic, including communications between agents, tools, and large language models (LLMs). This gateway is designed for deterministic enforcement, meaning it applies clear, non-negotiable rules outside the agent's own reasoning loop, preventing the agent from seeing or bypassing the controls.

On the other side is Rubrik Agent Cloud, which feeds intelligent policy signals into that enforcement gateway. The engine behind this is SAGE (Semantic AI Governance Engine), Rubrik's proprietary small language model (SLM) trained specifically for governance. Unlike static, keyword-based filters, SAGE interprets the intent behind natural language policies—such as "Do not share personally identifiable information with external tools"—and evaluates an agent's proposed actions in real-time. This semantic detection is probabilistic, capable of understanding nuance and context that rigid rules would miss. The integration's brilliance lies in how these two systems work together: Rubrik’s SAGE provides the nuanced, context-aware detection, and Amazon's AgentCore provides the final, deterministic allow-or-deny enforcement. This ensures that even if an agent finds a clever way to phrase a risky request, the gateway will block it based on the policy signal from SAGE.

Beyond Detection: The Power of Resilience and Remediation

While other security vendors are entering the AI security space with solutions for threat detection and prompt injection, Rubrik is leveraging its deep heritage in data protection and cyber resilience to offer a capability that few others can: remediation. The integration highlights a standout feature of Rubrik Agent Cloud known as 'Agent Rewind.'

This function moves beyond simply monitoring and blocking risky actions. It provides a complete, immutable audit trail of every action an AI agent takes. If an agent makes a mistake or acts maliciously—for instance, by deleting a critical database or corrupting production data—Agent Rewind allows administrators to instantly and precisely undo that specific action. It works by correlating the destructive agentic action with a previous healthy data snapshot from Rubrik's core backup and recovery platform, effectively making AI agent mistakes reversible. As one industry analyst noted, this is one of the first offerings to directly link visibility of AI agent actions with the tangible ability to undo them, transforming observability into genuine operational resilience.

This capability is a strategic game-changer. It provides a safety net that fundamentally de-risks the deployment of autonomous systems. For enterprises, this means they can empower agents to perform more meaningful, high-stakes tasks, knowing that a pathway to recovery is in place if something goes wrong. It’s a powerful assurance that helps shift the organizational mindset from fear of AI failure to confidence in AI experimentation.

Strategic Implications for the Cloud AI Landscape

This partnership is a powerful strategic maneuver that redefines the competitive landscape for cloud AI security. For Rubrik, which recently went public, it firmly establishes its position at the intersection of security and AI operations, integrating its advanced governance capabilities directly into one of the world's leading cloud AI ecosystems. This move provides immense validation and a direct channel to a vast market of enterprises building on AWS.

For AWS, the collaboration enriches the Amazon Bedrock platform with a best-of-breed security and resilience layer, making it a more compelling choice for highly regulated or risk-averse industries. By partnering with a specialist like Rubrik, AWS can offer its customers a comprehensive solution that addresses the full lifecycle of AI agent security—from policy creation and real-time enforcement to post-incident remediation. While the integration is still designated as "upcoming," its announcement sets a new standard for what enterprise-grade AI governance should entail. It signals to the market that a truly secure AI strategy requires a multi-layered approach that combines intelligent detection, deterministic control, and robust resilience, paving the way for a future where autonomous AI can be deployed not just at scale, but with a level of trust that businesses can depend on.