AI Red Team Tester — Remote, Flexible Contract

Alignerr Corp.; Indonesia; A distancia; IDR 377.149.000 - 1.005.731.000; A tiempo parcial

Descripción de la vacante

Alignerr is seeking AI Red Team Testers to challenge cutting-edge AI models, probing weaknesses, exposing blind spots, and helping teams build safer, more reliable systems. This is a fully remote, flexible hourly contract open to imaginative thinkers from any background.

No cybersecurity or technical experience required—just curiosity and a knack for thinking differently. Responsibilities include designing prompts, eliciting problematic outputs, documenting failure modes, and rating severity,

Formación

• This role relies on curiosity and lateral thinking to design prompts and scenarios.
• Ability to document failure modes with clear, reproducible steps and observations.
• Strong written communication to explain findings clearly and precisely.
• Systematic and detail-oriented approach to recording results.

Responsabilidades

• Design creative prompts, scenarios, and conversational strategies to probe AI model weaknesses
• Attempt to elicit incorrect, unsafe, biased, or otherwise problematic outputs from AI systems
• Document failure modes with clear, reproducible steps and structured observations
• Rate the severity and potential impact of discovered issues
• Collaborate asynchronously with AI safety and research teams to improve model quality
• Explore edge cases across a wide range of topics, tones, and contexts

Conocimientos

Curiosity

Creative thinking

Written communication

Detail-oriented

Independent work

Descripción del empleo

Alignerr is seeking AI Red Team Testers to challenge cutting-edge AI models, probing weaknesses, exposing blind spots, and helping teams build safer, more reliable systems. This is a fully remote, flexible hourly contract open to imaginative thinkers from any background.

No cybersecurity or technical experience required—just curiosity and a knack for thinking differently. Responsibilities include designing prompts, eliciting problematic outputs, documenting failure modes, and rating severity,