New niche research titles
Frontier AI labs are creating specialised, research‑oriented job titles—examples include 'Abuse Investigator' at OpenAI and roles like 'Research Scientist, Honesty' or 'Research Engineer, Universes' at Anthropic—signalling demand for skills beyond conventional ML engineering. These listings reflect increasingly granular role definitions that pair domain expertise (safety, honesty, abuse) with research mandates. (x.com/Hesamation/status/2043689341659004953)
Frontier artificial intelligence labs are posting narrower research jobs, splitting work once grouped under machine learning into titles like Abuse Investigator and Honesty researcher. (openai.com) (anthropic.com) OpenAI’s careers page listed 640 jobs on April 14, 2026, including Abuse Investigator, Abuse Investigator (CBRN), Abuse Investigator - Child Safety, and Technical Abuse Investigator under Intelligence & Investigations. (openai.com) Anthropic’s jobs board listed 428 roles on April 14, 2026, including Research Scientist / Engineer, Honesty, plus roles such as Research Engineer, Interpretability, Research Engineer, Frontier Red Team (Autonomy), and Privacy Research Engineer, Safeguards. (anthropic.com) These titles map to a basic shift in how labs describe the work: not just training models, but studying how people misuse them, how systems behave under pressure, and how to build controls that scale. OpenAI says its investigations team identifies “new types of abuse,” while Anthropic says it is building “reliable, interpretable, and steerable” systems. (openai.com) (anthropic.com) OpenAI’s postings show how specific that work has become. Its CBRN role focuses on biological and chemical threats, its child safety role covers attempts to sexually exploit minors, and its technical investigator role asks candidates to build SQL and Python tools that help other investigators detect harm faster. (openai.com 1) (openai.com 2) (openai.com 3) Anthropic’s listings show the same pattern on the research side. Alongside general model-building roles, the company is hiring for focused areas including honesty, interpretability, cybersecurity reinforcement learning, observability, and safeguards. (anthropic.com 1) (anthropic.com 2) (anthropic.com 3) The language in both companies’ recruiting pages points to hybrid hiring. OpenAI says safe deployment requires “a wide range of disciplines and backgrounds,” and Anthropic tells applicants that “engineers here do lots of research, and researchers do lots of engineering.” (openai.com) (anthropic.com) That makes these jobs different from a standard machine learning engineer opening. A child safety investigator is expected to analyze user behavior and support legal reporting, while a technical abuse investigator is expected to turn casework into reusable pipelines, dashboards, and internal tools. (openai.com 1) (openai.com 2) The result is a hiring map that looks more like a research hospital or intelligence unit than a software startup. The labs are still recruiting model builders, but their public job boards now carve out separate specialties for misuse, monitoring, honesty, red-teaming, and enforcement. (openai.com) (anthropic.com)
