Responsibilities
- Create and deploy safeguards for AI systems that use external tools, focusing on access restrictions and approved tool lists
- Ensure separation of context and memory between AI agent sessions to prevent cross-contamination
- Validate each action taken by an AI agent at the individual step level
- Apply security measures based on the OWASP Agentic AI Top 10, such as defending against prompt injection and data leaks
- Develop real-time enforcement systems like interceptors or proxies for monitoring tool and API calls
- Establish policy frameworks that make and enforce access decisions during AI execution
- Enforce limits on AI activity, including rate controls, execution duration, and emergency shutdown capabilities
- Block actions initiated by AI systems that are unauthorized or potentially harmful
- Design identity and access management for automated agents with temporary credentials and narrowly defined permissions
- Distinguish clearly between access granted to humans and non-human entities
- Tightly bind agent identity to the specific task and execution environment
- Ensure all actions taken by AI systems can be traced and reviewed
- Implement comprehensive logging and tracing for AI inputs, tool usage, and decision paths
- Develop detection systems using behavioral baselines and anomaly detection
- Detect and flag unusual patterns in tool usage by AI agents
- Identify and report on suspicious structures or content in AI prompts
- Monitor for and alert on unexpected or unauthorized data access attempts
- Conduct threat modeling for agentic systems using the MAESTRO framework, including mapping capabilities and trust boundaries
- Model potential misuse and adversarial behaviors in AI systems
- Turn threat modeling results into actionable security controls and detection rules
- Protect developers using AI tools by preventing exposure of sensitive data
- Verify the safety and accuracy of code and actions generated by AI tools
- Limit automation capabilities to prevent unsafe or uncontrolled behavior
- Support secure use of AI-powered coding tools with embedded validation layers
- Promote secure prompting and output handling practices in AI-assisted development environments
Work Arrangement
Remote (Worldwide)
Other
- We are proud to be an Equal Opportunity Employer
- Diversity, equity, and inclusion are central to our values
- We are committed to fairness and respect toward our customers, partners, and employees
- We aim to build a diverse workforce to better serve customers across more than 175 countries
- Continuous learning and personal growth are integral to our culture
- We welcome applications even if candidates do not meet every listed requirement
- We strive to create an environment where people can be themselves and do their best work
