Safeguards ML Engineer: Misuse Detection & Safe AI

Anthropic Limited; San Francisco (CA); On-site; USD 350,000 - 500,000; Full time

Job summary

Anthropic Limited is seeking ML/Research Engineers to help detect and mitigate misuse of AI systems within the Safeguards ML team. You will build scalable detectors, monitor harms across contexts, and develop defenses to keep products safe as capabilities advance.

Strong emphasis on responsible AI and collaboration across research and deployment teams. The role requires 4+ years in ML/AI research or engineering, proficiency in Python, and the ability to move from experiments to production.

Qualifications

• 4+ years of experience in ML engineering, research engineering, or applied research in academia or industry.
• Proficiency in Python and experience building ML systems.
• Comfortable with the research-to-deployment pipeline from experiments to production.
• Concern for misuse risks of AI and a drive to mitigate them.
• Strong communication skills and ability to explain complex concepts to non-technical stakeholders.

Responsibilities

• Develop classifiers to detect misuse and anomalous behavior at scale, including synthetic data pipelines for training.
• Build systems to monitor harms across exchanges and analyze signals across contexts.
• Evaluate and improve safety of agentic products, including threat models and mitigations for prompt injection.
• Conduct research on automated red-teaming, adversarial robustness, and related misuse testing.

Skills

ML engineering

Python

Deployment pipeline

Communication

Misuse mitigation

Job description

Anthropic Limited is seeking ML/Research Engineers to help detect and mitigate misuse of AI systems within the Safeguards ML team. You will build scalable detectors, monitor harms across contexts, and develop defenses to keep products safe as capabilities advance.

Strong emphasis on responsible AI and collaboration across research and deployment teams. The role requires 4+ years in ML/AI research or engineering, proficiency in Python, and the ability to move from experiments to production.