Red Team Engineer, Safeguards

About this role

Anthropic seeks a Red Team Engineer to identify vulnerabilities in its AI systems and products through adversarial testing. You'll develop creative attack scenarios, research novel exploitation methods for emerging AI capabilities, and collaborate with teams to implement safety improvements before systems are exploited.

What you'll do

• Conduct adversarial testing across product surfaces, combining multiple exploitation techniques into sophisticated attack chains
• Research and implement testing approaches for emerging capabilities like agent systems and tool use
• Design full kill-chain attacks that simulate real-world threat actors targeting specific malicious objectives
• Build automated testing frameworks and systematic methodologies for continuous safety assessment at scale
• Establish metrics to measure detection effectiveness against novel abuse patterns
• Partner with Product, Engineering, and Policy teams to translate security findings into concrete improvements

What they're looking for

• Penetration testing and red teaming
• Model jailbreaking and prompt injection testing
• Web application security (Burp Suite, Metasploit)
• Custom automation and scripting for security testing
• LLM-specific testing framework development
• API security and rate-limiting analysis
• Business logic vulnerability testing
• AI/ML adversarial security

Benefits

• Annual salary: $320,000–$405,000 USD
• Remote-friendly work arrangement with 25% minimum office time
• Work on cutting-edge AI safety challenges
• Collaboration with world-class researchers and engineers