SentinelOne said its Prompt Security capabilities will integrate with Amazon Bedrock AgentCore to help enterprises apply runtime controls to AI agents, while AWS said AgentCore policies can use inputs from security providers for deterministic enforcement of agent actions.
SentinelOne announced that its Prompt Security capabilities will integrate with Amazon Bedrock AgentCore to provide runtime guardrails for enterprise AI agents.
In a company announcement, SentinelOne said the integration is intended to help organizations monitor and control AI agents while they are operating. The company said its Prompt Security capabilities are designed to detect prompt-injection attempts, identify potential personally identifiable information exposure, validate tool use, and help prevent data leakage.
SentinelOne described the work as an integration with Amazon Bedrock AgentCore, AWS’s service for building and operating AI agents. According to SentinelOne, the goal is to add security checks around agent behavior at runtime, rather than relying only on controls applied before deployment.
AWS, in a blog post about new Amazon Bedrock AgentCore capabilities, said customers will be able to use inputs from security providers, including SentinelOne, in AgentCore policies. AWS said AgentCore can apply deterministic gateway enforcement for agent actions, meaning policy decisions can be enforced through defined controls when an agent attempts to use tools or take actions.
That approach is relevant because AI agents often operate by receiving instructions, calling tools, retrieving information, and taking steps across enterprise systems. AWS’s description indicates that third-party security inputs can become part of the rules that decide whether a requested action should proceed.
The main risks highlighted by SentinelOne include prompt injection, PII exposure, unsafe tool use, and data leakage. Prompt injection refers to attempts to manipulate an AI system’s instructions or behavior through crafted inputs. PII exposure concerns the handling or disclosure of personal data. Tool-use validation addresses whether an agent’s requested action is appropriate under enterprise policy.
The Fly, published via TipRanks, also reported SentinelOne’s upcoming Amazon Bedrock AgentCore integration and summarized the company’s Prompt Security detections that will feed into AgentCore’s policy engine.
The available announcements describe an integration between SentinelOne Prompt Security and Amazon Bedrock AgentCore, with SentinelOne providing security-related detections and AWS providing policy enforcement mechanisms inside AgentCore. The sources do not provide detailed pricing, deployment timelines, or customer adoption figures.
For enterprises experimenting with AI agents, the announcement points to a developing pattern: runtime policy enforcement combined with specialized security monitoring. SentinelOne is positioning Prompt Security as one of the security inputs for that model, while AWS is presenting AgentCore as the environment where those policies can be applied to agent actions.
SentinelOne announced that its Prompt Security capabilities will integrate with Amazon Bedrock AgentCore to provide runtime guardrails for enterprise AI agents.
SentinelOne outlines AI agent protections In a company announcement, SentinelOne said the integration is intended to help organizations monitor and control AI agents while they are operating.
The company said its Prompt Security capabilities are designed to detect prompt injection attempts, identify potential personally identifiable information exposure, validate tool use, and help prevent data leakage.
Continue reading
