Full Report
AI agents are creating a new digital workforce, leaving IT teams struggling with a rapidly increasing number of identities to secure.
Analysis Summary
# Cybersecurity Risks Associated with AI Agent Deployment
## Key Points
- The deployment of AI agents exponentially increases the number of identities that cybersecurity teams must secure, as each agent acts as a new entity requiring credentials and access control.
- This proliferation of identities (for individuals, machines, software, and now AI agents) is straining existing identity management infrastructures and directories.
- Cybercriminals are paying close attention to how AI agents are built and deployed, anticipating new, rich attack surfaces.
- Failure to secure credentials for AI agents compounds existing issues where stolen credentials are used for unauthorized access without resorting to malware.
## Threat Actors
- Cybercriminals (general threat) are anticipated to leverage the increased number of credentials associated with AI agents for unauthorized access.
- No specific named threat groups were identified in connection with launching attacks via AI agents, but their interest in this new attack surface is noted.
## TTPs
- **Credential Exploitation:** Leveraging stolen or easily accessible credentials associated with AI agents to log into various applications and services.
- **Attack Surface Expansion:** Exploiting vulnerabilities related to the new identities assigned to AI agents.
- **Interaction Exploitation:** The ability of AI agents to interact with other agents and humans presents unique vectors for exploitation, though specific methods are not detailed.
## Affected Systems
- Organizational IT environments deploying AI agents.
- Existing identity directories and authentication platforms, which may be overwhelmed by the scale of new identities.
## Mitigations
- Aggressively implement **Zero Trust IT policies** to manage the heightened identity management challenge.
- Focus efforts on **identity management**, as it is core to enforcing Zero Trust policies across human and machine (AI agent) identities.
- Cybersecurity teams must immediately begin determining how best to secure the deployed AI agents, recognizing that no standard framework is universally established yet.
- Consider leveraging existing or proposed security frameworks for securing AI agents, such as: MAESTRO, STRIDE, PASTA, LINDUNN, OCTAVE, TRIKE, and VAST.
## Conclusion
The rapid rollout of AI agents presents a critical security challenge by massively increasing the enterprise attack surface through identity proliferation. Organizations must proactively address AI agent security now, predominantly through the comprehensive adoption of Zero Trust principles centered around robust identity management, before existing security infrastructures are overwhelmed by the scale of new machine identities.