Amazon agentic AI attack — Jul 2025
A hacker compromised Amazon Q by injecting malicious commands into its GitHub repository which were subsequently included in a public software release
- Reported date
- Jul 23, 2025
- Target
- Amazon
- Agent type
- Coding agent
- Agent role
- Compromised agent
The exact incident date was not established. This entry is dated by its source report.
A security breach involving Amazon’s AI coding assistant, Q, resulted in the inclusion of unauthorized, potentially destructive commands within a public software update. The incident highlights vulnerabilities in the development pipeline for AI-powered tools.
What happened
According to 404 Media, a hacker successfully compromised a version of Amazon Q by submitting a malicious pull request to the tool’s GitHub repository. The injected code contained a prompt instructing the AI agent to act as a system cleaner with access to filesystem tools and bash, specifically tasking it to "clean a system to a near-factory state and delete file-system and cloud resources."
Amazon subsequently included this unauthorized update in a public release of the assistant. While the hacker claimed they could have caused significant damage through this access, the actual risk of the injected code successfully wiping user computers appears to be low. The incident serves as an example of how attackers are increasingly targeting the repositories and development workflows of AI-powered coding assistants to gain unauthorized influence over software behavior.
Evidence in the reporting
- Incident evidence
- Amazon included the unauthorized update in a public release
- Agent involvement
- told the software to wipe users’ computers