AI Safety Evaluator

Job Description

The AI Safety Evaluator is responsible for assessing whether artificial intelligence systems operate within established safety requirements and produce responses that are appropriate, responsible, and aligned with applicable usage guidelines. The role involves reviewing AI-generated content, evaluating responses to sensitive or potentially harmful requests, and identifying situations in which models may provide unsafe instructions, inappropriate recommendations, or misleading information that could create risks for users.

The successful candidate will examine how AI systems respond to a wide range of requests, including sensitive discussions, potentially dangerous activities, cybersecurity-related questions, privacy concerns, and other scenarios requiring careful judgement. They will assess whether the system appropriately recognises risks, applies relevant safeguards, communicates limitations, and provides safe alternatives where necessary.

The evaluator will also examine whether safety controls function consistently across different prompts and contexts. This may involve testing variations of the same request, reviewing model responses under challenging conditions, and documenting cases where safeguards are ineffective or unnecessarily restrictive.

The role is important in helping AI development teams strengthen responsible AI practices, reduce harmful outputs, and improve the consistency of safety-related model behaviour.

Key Responsibilities:

Evaluate AI responses for compliance with established safety standards.
Identify harmful, inappropriate, misleading, or potentially dangerous outputs.
Review how AI systems respond to sensitive and high-risk requests.
Assess whether safety safeguards are applied consistently.
Identify situations where AI models fail to recognise potentially harmful instructions.
Evaluate whether refusals and safety explanations are appropriate and clear.
Review privacy-related responses and potential exposure of sensitive information.
Document safety failures and recommend corrective improvements.
Participate in structured safety testing and evaluation exercises.
Prepare detailed reports for AI research, policy, and quality assurance teams.
Qualifications and Requirements:

Bachelor’s degree in Computer Science, Cybersecurity, Information Technology, Psychology, Law, Public Policy, or a related discipline is preferred.
Strong analytical and ethical judgement.
Understanding of responsible AI principles and digital safety practices.
Ability to identify harmful or potentially unsafe AI-generated content.
Strong written communication and reporting skills.
Familiarity with AI evaluation frameworks and content safety guidelines is an advantage.
Ability to handle sensitive evaluation material professionally.
Strong attention to detail and consistency.