AI tester with French language

About the role

This position serves as a critical evaluation function within the development lifecycle of an AI system that is currently under construction. Your primary ownership is the systematic probing of the agent to identify how it behaves when explicitly encouraged to generate harmful, offensive, or dangerous material. You will own the definition of security boundaries by designing and executing test scenarios that push the model beyond its intended limits. Every interaction you conduct will be analyzed to determine whether the system's safety mechanisms fail under adversarial pressure. You are responsible for maintaining rigorous records of each test case, including the input prompt and the resulting model output. Your analysis will directly influence whether the system is deemed safe enough for release to the general public. Success in this role means you contribute to preventing harmful content from ever reaching end users. The work is conducted remotely on a full-time basis for a contract duration of several months, with the timeline being dependent on project milestones and the validity of the findings.

Key facts

What you'll do

• Design and execute complex question sets specifically intended to challenge the AI agent across predefined safety and harm categories.
• Construct multi-stage prompts that methodically attempt to bypass established safety protocols and content filters.
• Test whether the system is capable of generating toxic, offensive, or otherwise dangerous language when subjected to adversarial input.
• Classify every agent response into safe or risky categories, ensuring a clear separation between acceptable and unacceptable outputs.
• Monitor entire conversation threads to identify behavioral patterns and subtle inconsistencies in the model's safety performance.
• Translate raw test observations and qualitative findings into actionable adjustments for the training data and model configurations.
• Conduct verification testing cycles after modifications to confirm that harmful content is effectively reduced or eliminated.
• Maintain precise written documentation for every test scenario, including hypotheses, procedures, and observed outcomes.
• Communicate detailed results and logical conclusions to remote technical teams with precision and clarity.
• Validate findings through repeated testing to ensure issues are reproducible and not the result of transient anomalies.

Requirements

• Possess authorization to engage in remote full-time employment for a duration spanning several consecutive months.
• Demonstrate concrete experience in AI testing, evaluation projects, or similar technical assurance roles.
• Bring a background in red teaming, security research, or adversarial testing methodologies is a strong eligibility factor.
• Exhibit fluency in French language to the level required for creating complex and nuanced test prompts.
• Maintain a strong grasp of English to read internal documentation and communicate effectively with global teams.
• Ensure you have stable and high-speed internet access at all times to support uninterrupted testing activities.
• Provide a dedicated quiet workspace that allows for deep concentration and meticulous attention to detail.
• Commit to consistent daily availability to ensure continuity in the testing schedule and timely progress reporting.
• Strictly adhere to all safety protocols when handling, reviewing, or documenting toxic or harmful material.
• Equip yourself with suitable computer hardware and software necessary to run the testing environment and tools.

Nice to have

• Bring prior experience specifically in AI safety testing frameworks and methodologies.
• Apply knowledge of red teaming strategies to anticipate model weaknesses and exploit them constructively.
• Understand common prompt injection techniques to design more sophisticated and effective test cases.
• Leverage this specialized knowledge to create test scenarios that are more challenging and revealing.

Practical notes

No specific tools are prescribed for this role, as all necessary resources will be determined during the onboarding process. You should review the official application page for precise details regarding the application mechanism and logistical information. Confirming instructions and any additional context regarding the project timeline are strongly recommended by consulting that official page.