Maze is a cloud vulnerability management platform that uses AI agents to investigate and resolve security vulnerabilities. The system replicates security engineer workflows to analyze vulnerabilities within the context of cloud environments, compensating controls, and business requirements. The platform investigates exploitation requirements for each vulnerability by checking for specific conditions such as software versions, system configurations, and active services. It evaluates technical factors including weaponized exploits, network access controls, WAF configurations, and IAM permissions alongside business context like system confidentiality ratings and customer-facing applications. Maze filters out non-exploitable vulnerabilities by determining whether required exploitation conditions are present in the environment. The system generates investigation reports that can be used for audit purposes. For exploitable vulnerabilities, it prioritizes based on likelihood and impact assessment. The platform automates remediation through intelligent workflows that create incidents, deploy security controls, generate tickets, and send notifications. It operates in an agentless manner without requiring software installation on target machines. Deployment requires a read-only role in cloud environments and API connections to vulnerability scanners. Each customer receives a single-tenant environment hosted on AWS in either US or Europe regions. The system uses language models from providers like AWS Bedrock and Google that do not share customer data for training purposes.