According to a recent LinkedIn post from Glow, research by its Glow Labs unit identified two AWS security-scanning integrations that could expose source code and secrets to attackers. The post names the aws-agents-for-devsecops plugin and AWS Security Agent MCP server, and says AWS published fixes on Sept. 10 for CVE-2026-87912 and CVE-2026-87913.
According to the post, both tools uploaded scanned code to Amazon S3 using bucket names derived from an AWS account ID and region without verifying ownership. It says an attacker who knew the public account ID could register the bucket first, potentially receiving archives containing .env files, private keys, or Terraform state.
The research also described the missing ownership check as residing in a Markdown skill file rather than application code, and warned that upgrading may not release a bucket name already registered by an attacker. For investors, the issue highlights the difficulty of securing AI-enabled development tools and the potential for overlooked configuration risks to create data-exposure liabilities.
The findings could support demand for broader AI-tool discovery, code scanning, and cloud-configuration monitoring, areas that may strengthen Glow Labs’ positioning in cybersecurity. However, the LinkedIn post does not quantify customer exposure, remediation costs, or commercial impact, so its effect on Glow’s financial outlook remains uncertain without independent validation and further disclosure.


