.jpg)
AI agents are moving into production fast. They read email, call tools, query databases, and trigger business processes, often chaining those steps in ways no single team designed. Building them has never been easier. Securing what happens after an agent is manipulated is the hard part.
That is the gap Salt and AWS will tackle together on Tuesday, November 10, at 9:00 am PT (12:00 pm ET) in a live session: AWS + Salt: Build Fast, Stay Secure with Agentic AI.
Agents are acting, and most teams can't see the full path
Salt's latest research, a survey of more than 380 organizations, shows how far deployment has outrun governance:
- 86.4% of organizations have at least one AI agent running, and more than half run ten or more.
- 49.8% confirmed or suspected that an agent took an action in the past year that nobody intended or authorized.
- 12.5% can consistently trace a complete agent path, from the initial instruction through tools and APIs to the final outcome.
- 54.3% have confirmed or suspected an unauthorized, shadow, or misconfigured MCP server.
Salt Labs research from September showed what this looks like in practice: a single malicious email was enough to hijack an agentic AI platform and reach a user's connected email, cloud storage, and code repositories. No stolen password or clicked link was needed. The platform's own guardrail detected the attack, and it still couldn't stop the code from running.
AWS provides the foundation. Salt follows the attack across it.
AWS gives teams the building blocks to build agentic applications on a secure footing. Amazon Bedrock, AgentCore, Guardrails, IAM, KMS, and CloudTrail cover infrastructure, identity, data, and governance from day one.
But once agents are live, security teams face a new question: what happens when an attacker manipulates the model and uses the agent to reach downstream tools and APIs? Model-level controls alone can't show you what an agent can access or what it does next.
That's where Salt comes in. The Salt Agentic Security Platform, now with native AI Detection and Response (AI-DR), connects a prompt injection to the compromised agent, the MCP tool it calls, the backend API behind it, and the business data at risk. One attack shows up as one story instead of five disconnected alerts.
What you'll learn
In this session, Nick Rago, VP of Product Strategy at Salt Security, and CJ Sturgess, Partner Solutions Architect at AWS, will show you how to:
- Apply the AWS AI Security Framework (use cases, layers, and phases) as agents gain autonomy.
- See what model-layer controls miss, including what an agent can reach and what it does next.
- Trace an attack from prompt to MCP server to API, and see why so few organizations can do this today.
- Extend runtime detection and response across Bedrock-built and homegrown agents without ripping out the security investments you already have.
You'll leave with a clearer picture of where your agentic attack surface really sits, and a practical way to extend protection across it.
Who should attend
Security leaders, cloud and platform architects, AppSec and API security teams, and anyone responsible for AI agents moving from pilot to production on AWS.
Reserve your spot
Tuesday, November 10 | 9:00 am PT / 12:00 pm ET
AWS + Salt: Build Fast, Stay Secure with Agentic AI
Register now to save your spot and get the join link →
Can't make it live? Register anyway, and we'll make sure you're set up for follow-up.
Want the full data behind these numbers? Download the State of Agentic AI and API Security 2H 2026 report.