ApplyBlast uses AI to match you with the right jobs, tailor your resume and cover letter, and apply automatically so you can land your dream job faster.

Features

Auto Apply Resume Optimizer Cover Letter Optimizer Interview Prep

Company

Social Media

Need Help?

team@applyblast.com

Enforcement Analyst, Agentic Alignment at Anthropic - ApplyBlast

Enforcement Analyst, Agentic Alignment

Anthropic

2 days ago

3.7 on Glassdoor

Boston, Massachusetts, United States

Hybrid

Full Time

Medior

0 applicants

View Job Listing

Apply to 100+ jobs

About this role

The Enforcement Analyst, Agentic Alignment role at Anthropic focuses on identifying and investigating harmful or unexpected behavior by AI models to ensure safe and beneficial outcomes. The position involves making enforcement decisions under the Usage Policy, assessing cyber-related agentic activity, and collaborating with cross-functional teams to improve policies and model behavior. This role requires strong analytical skills, experience in trust and safety or investigations, and the ability to work with technical data and AI systems.

Responsibilities

Review and investigate potential instances of Agentic Misalignment using transcripts and code
Make enforcement decisions under the Usage Policy and document reasoning behind each decision
Assess cyber-related agentic activity and determine its real-world harm potential

Skills

About Anthropic

anthropic.com

Company Size

51-200 employees

View Profile

Anthropic is an AI safety and research company focused on building reliable, interpretable, and steerable AI systems. It develops large language models (branded as Claude) and offers APIs and enterprise products that let organizations integrate conversational AI with safety-focused controls, moderation, and privacy features. The company prioritizes interpretability and alignment research, publishes technical work, and engages with policymakers to reduce risks from advanced AI. Customers choose Anthropic for its safety-first approach, controllability tools, and research-driven models.