Red Team Engineer, Safeguards

Anthropic is hiring a Red Team Engineer, Safeguards to strengthen the security posture of our AI products. Based in San Francisco with remote flexibility and periodic office attendance required, this full-time role focuses on identifying and neutralizing vulnerabilities in our AI systems before hostile actors can weaponize them.

About This Position

You'll adopt an offensive security mindset to uncover exploitable gaps across our AI platform ecosystem. This role extends beyond traditional cybersecurity into the unique threat landscape surrounding advanced AI: emerging capability-driven risks, abuse of AI-specific features, and coordinated multi-step attacks that chain together vulnerabilities. You'll reverse-engineer sophisticated threat actors' thinking to design realistic end-to-end attack scenarios, and you'll pioneer testing approaches for cutting-edge AI technologies that existing security frameworks weren't designed to evaluate.

Core Duties

• Conduct layered penetration tests across our product landscape by orchestrating multiple attack techniques into cohesive exploitation chains
• Research and operationalize novel security testing methods for emerging AI capabilities including agent systems, tool integrations, and new user interaction paradigms
• Build realistic attack simulations that mirror how determined threat actors pursue specific malicious objectives through your infrastructure
• Establish comprehensive, systematic testing protocols that measure security posture across the entire attack surface
• Engineer automated security assessment systems enabling continuous vulnerability detection at scale
• Partner with Product, Engineering, and Policy teams to convert security findings into practical remediation and system resilience improvements
• Develop quantitative frameworks for tracking detection rates and prevention effectiveness against novel abuse patterns

Qualifications

• Professional background in penetration testing, red team operations, or offensive security engineering
• Proven expertise in model jailbreaking and identification of subtle prompt injection vulnerabilities within large agentic systems
• Strong web application security foundation with hands-on proficiency in security testing tools and rapid custom automation development
• Experience designing and implementing AI-focused or LLM-specific testing automation and exploitation frameworks
• Demonstrated ability to discover novel attack vectors and synthesize multiple vulnerability chains into creative, high-impact exploits
• Published security research: issued CVEs, responsible disclosure reports, security research articles, or acknowledged bug bounty submissions
• Superior written and verbal communication skills enabling you to explain technical security concepts to audiences at all levels

About Anthropic

Anthropic develops AI systems engineered for reliability, transparency, and alignment with human values. We're dedicated to ensuring AI becomes a positive force for users and society. Operating as an integrated research organization, we concentrate resources on a handful of transformational initiatives rather than scattered projects. We treat AI research as fundamental science—drawing insights from physics and biology alongside computer science. Our culture prioritizes substantive collaboration and clear communication, with particular value placed on rigorous thinking about AI safety, alignment, and long-term societal implications.

Next Steps

To apply, complete your application directly on this page, or you'll be redirected to the employer's application platform to finish submitting there.