One prompt could expose every AWS Bedrock AgentCore agent in a region

Feeds.4Sysops •IT News • October 8, 2026

A single prompt to a public-facing Amazon Bedrock AgentCore agent could let an attacker take over other agents in the same AWS account and region, Zenity Labs researchers found. Their demonstration exposed credentials, source code, and private conversations; AWS has tightened default settings, but Zenity recommends custom, least-privilege roles. Source

The full story

This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.

Vulnerability in AWS Bedrock AgentCore Exposes AI Agents to Hijacking Researchers from Zenity Labs discovered a vulnerability in AWS Bedrock AgentCore that allowed an attacker to take control of all AI agents within the same AWS account and region using a single prompt to a public-facing agent. The flaw, termed 'AgentCorruption', exploited inadequate isolation of agents and broad default permissions, enabling access to sensitive data such as internal AWS credentials, private conversations, and source code. AWS has partially addressed the issue by tightening default execution roles and restricting access to internal metadata, but experts recommend further manual role assignments for security. The vulnerability highlights long-standing issues with AWS's Instance Metadata Service (IMDS) and its implications for cloud security. The research indicates that attackers could manipulate agent memory and exfiltrate credentials from AWS Secrets Manager, posing a significant risk to organizations relying on AI agents. 6 sources 70

Extracted Entities

One prompt could expose every AWS Bedrock AgentCore agent in a region

The full story

Extracted Entities

Attack Types (1)

CWE Weaknesses (1)

Platforms (1)