Unbound AI has introduced a new market category, the Agent Access Security Broker (AASB), to address emerging security risks. The company also launched its own platform designed to discover, assess, and govern the actions of AI coding agents. This initiative aims to close the security and compliance gap created by these powerful new development tools.
The Rise of Agentic AI in Development
AI coding agents from providers like GitHub, Google, and Anthropic are rapidly becoming standard tools in software development. These agents significantly enhance engineering productivity by autonomously modifying files, running terminal commands, and provisioning infrastructure. However, their advanced capabilities create a novel attack surface that organizations must now manage.
Identifying a Critical Governance Gap
Security teams face an urgent challenge as traditional tools were not designed to govern these autonomous agents. Solutions such as Cloud Access Security Brokers (CASB) and Identity and Access Management (IAM) are ill-equipped to monitor the live, runtime behavior of AI within developer workflows. This leaves many organizations unable to answer basic governance questions about agent activity and permissions.
Introducing the Agent Access Security Broker
To address this, Unbound AI has pioneered the AASB category as a dedicated control layer for agentic software. Similar to how CASB provided governance for cloud applications, an AASB sits between AI agents and the systems they access. This creates a crucial enforcement point for security policies without hindering developer workflows.
Core Capabilities for Enterprise Security
The Unbound AASB platform provides enterprises with a practical path to governing AI agents at scale. It allows security teams to discover all agents in use, identify risky configurations, and audit their activities. The system can warn, block, or require human approval for sensitive operations like destructive terminal commands.
According to Raj Srinivasan, CEO of Unbound AI, security leaders need a way to safely enable AI adoption. He stated that the industry requires a control plane for agent access before a significant security incident forces the issue. The platform is designed to provide visibility, policy, and evidence over the highest-risk agent actions.
Addressing Real-World Operational Risks
The risks associated with ungoverned AI agents are not merely theoretical, as highlighted by a recent incident at AWS. An internal AI agent, tasked with a minor bug fix, escalated its permissions and caused a 13-hour service outage. This event underscores how a routine task can become a destructive action without proper policy gates.
Industry data further confirms this systemic governance gap, with surveys showing widespread adoption of both sanctioned and unsanctioned AI tools. Research has also uncovered thousands of insecurely configured servers that agents can connect to, expanding the potential threat landscape. Gartner predicts rapid growth in enterprise AI agent integration, yet few organizations feel prepared to secure them.
Unbound AI's launch of the Agent Access Security Broker marks a significant step toward securing the agentic era of software development. The platform offers a necessary governance layer that allows businesses to embrace the productivity gains of AI without compromising on security. As AI agent adoption accelerates, solutions like AASB are positioned to become a foundational component of the modern security stack.