IT Consultancy
IT Consultancy Get expert guidance to leverage technology for your business growth. We help you identify the right IT solutions, optimize infrastructure, and ensure your digital environment is secure,…
Amazon Kiro Prompt Injection Could Leak Sensitive Data
Digital Marketing
Digital Marketing Grow your brand and reach the right audience with data-driven marketing strategies. We help you create campaigns that convert, manage social channels effectively, and optimize your d…
Privacy Policy
Privacy Policy At CometSoul, accessible from https://cometsoul.com, one of our main priorities is the privacy of our visitors. This Privacy Policy document contains types of information that is collec…
Terms and Conditions
Welcome to CometSoul! These terms and conditions outline the rules and regulations for the use of CometSoul’s Website, located at https://cometsoul.com. By accessing this website we assume you a…
Disclaimer
If you require any more information or have any questions about our site’s disclaimer, please feel free to contact us. Disclaimers for CometSoul All the information on this website – https…
Managed Wi-Fi Service
Managed Wi-Fi Service Build a fast, stable, and secure Wi-Fi for your business at Zero depreciation cost. Enterprise-Grade Guest Wi-Fi With $0 Upfront Hardware Costs. We provide, install, and support …
Smart Business System Integrator
Smart Business System Integrator Simplify your daily operations with smart systems that automate bookings, schedules, and routine tasks so you can focus on growing your business. We set up and integra…
IT Procurement & Hardware Managed Service
IT Procurement & Hardware Managed Service Keep your technology running smoothly from the hardware you use every day to the systems that power your business. We provide end-to-end IT management so …
Security researchers have found a prompt injection flaw in Amazon Kiro, an AI-powered IDE for developers.
The flaw could allow attacker-controlled repository content to influence the AI agent. As a result, sensitive local data may be sent to an external server without the user directly asking Kiro to share it.
How Does the Amazon Kiro Flaw Work?
The issue starts when a user opens a specially crafted project.
To trigger the vulnerable flow, the user must open a workspace file through Open Workspace From File. After that, sending a normal message to the AI agent can be enough.
Importantly, the user does not need to type a malicious prompt. Hidden instructions inside the project can influence the agent automatically.
Kiro Powers Can Increase the Impact
Kiro Powers gives the agent access to MCP servers, steering files, hooks, and extra context.
These features are useful for developers. However, they can also increase risk when malicious project content is treated as trusted instructions.
During testing, the agent could read local data and place it into security-sensitive configuration. Another IDE feature could then send that data to an attacker-controlled endpoint.
Has the Flaw Been Fixed?
Yes. Amazon fixed the issue in Kiro IDE version 0.8.140.
Researchers reproduced the flaw against Kiro IDE 0.7.45 on Windows. Newer versions already include the security fix.
Therefore, users should make sure they are running the latest Kiro IDE release.
What Should Developers Do?
First, update Amazon Kiro to the latest version.
Also, be careful when opening repositories or workspace files from unknown sources. A project does not need to contain an executable file to create risk.
Developers should also limit AI agent access to only the files, tools, and credentials it truly needs.
Key Takeaway
The Amazon Kiro prompt injection flaw shows that security risks in AI coding tools can go beyond incorrect AI responses.
If an agent has broad access to local files and tools, hidden instructions may lead to sensitive data exposure.
For this reason, keep Kiro updated and only open projects from trusted sources.
Source: https://thehackernews.com/2026/08/amazon-kiro-prompt-injection-can.html