Your tasks will involve writing adversarial; prompts to identify weaknesses in various AI models, including Large Language Models (LLMs), Text-to-Image, Text-to-Video, and beyond. You'll also oversee data management to guarantee the highest quality of outputs.
Responsibilities:
Developing adversarial and risky prompt strategies across several areas of abuse to expose potential vulnerabilities in models.
Managing projects end-to-end, from initial planning and oversight through quality assurance to final delivery.
Handling extensive datasets across multiple languages and areas of abuse, ensuring precision and meticulous attention to detail.
Ongoing investigation into new tactics for circumventing foundational models' safety measures.
Working alongside diverse teams, engineering, product, policy, to tackle new challenges and craft forward-thinking strategies and resolutions.
Promoting a culture of knowledge exchange and continual learning within the team.
Must have:
Background in AI Safety and/or Responsible AI and/or AI Ethics
Familiarity with recent Generative AI models and agents is essential, though direct technical experience is not a prerequisite.
Command of English at a near-native level.
Attention to detail, organizational capabilities, and the capacity to juggle numerous tasks concurrently.
Additional Wants:
Experience with various model types (Text-to-Text, Text-to-Image) is desirable.
Prior experience with OSINT (Open Source Intelligence) will be considered an asset.
A self-starter attitude, with the energy to excel in a fast-moving and variable environment.
Alice is a trust, safety, and security company built for the AI era. We safeguard the communicative technologies people use to create, collaborate, and interact—whether with each other or with machines.
In a world where AI has fundamentally changed the nature of risk, Alice provides end-to-end coverage across the entire AI lifecycle. We support frontier model labs, enterprises, and UGC platforms with a comprehensive suite of solutions: from model hardening evaluations and pre-deployment red-teaming to runtime guardrails and ongoing drift detection.
