Agentic Alignment Enforcement Analyst

Anthropic Limited; United States; Remote; USD 245,000 - 285,000; Full time

Job summary

Anthropic Limited is seeking an Enforcement Analyst focused on Agentic Alignment to identify and investigate harmful model behavior. You will work with Policy, Threat Intel, Engineering, Research, and Transparency teams to sustain safe and beneficial AI as Claude scales usage.

Expect on-call duties and exposure to sensitive content. You will review transcripts, analyze code, and write concise case write-ups to support enforcement decisions under the Usage Policy.

Qualifications

• Background in security, cyber threat intelligence, or a similar technical field.
• Experience in trust and safety, enforcement, investigations or related field.
• Structured investigations from first signal to final finding with defensible written findings.

Responsibilities

• Review and investigate agentic misalignment cases using transcripts, tool traces, and code.
• Triage cases by severity and urgency and document the reasoning behind decisions.
• Translate detection outputs into scalable review workflows across surfaces and teams.
• Assess cyber-related agentic activity to determine real-world harm potential.
• Write clear, well-evidenced case write-ups for stakeholders from enforcement to leadership.
• Collaborate with Threat Intelligence, Research, Engineering and Policy partners.

Skills

Agentic misalignment

Enforcement decisions

Transcript analysis

Code review

Cross-functional collaboration

Education

Bachelor's degree

Tools

Claude

Job description

Anthropic Limited is seeking an Enforcement Analyst focused on Agentic Alignment to identify and investigate harmful model behavior. You will work with Policy, Threat Intel, Engineering, Research, and Transparency teams to sustain safe and beneficial AI as Claude scales usage.

Expect on-call duties and exposure to sensitive content. You will review transcripts, analyze code, and write concise case write-ups to support enforcement decisions under the Usage Policy.