AI Data Annotator, Cybersecurity

About ProtocolClear

ProtocolClear is an independent consultancy led by its founder, focused on security, privacy, AI governance, and compliance. The company runs an in-house annotation and security team in Lagos to produce high-quality human judgments for AI training. This hiring need exists because the team requires annotators with genuine cybersecurity knowledge to label and evaluate security-related data accurately.

Role overview

This is a specialist annotation role centered on security, privacy, and governance content used to train and evaluate AI systems. The work involves checking whether model answers to security questions are technically correct, classifying security-related material, making RLHF-style preference and quality judgments on cybersecurity, privacy, and governance outputs, applying harm and safety labels in a security context, and working with adversarial examples such as prompt injection and jailbreaks. The founder, who is an AI security architect, provides the rubric, and you are expected to apply real cybersecurity expertise consistently and defensibly.

This stream is starting with a rubric-based setup that comes with higher per-task pay and longer estimated task times, and it will be adjusted once the client confirms the final terms.

Quality expectations

Your work will be reviewed to a five-star client standard, and accuracy matters more than style. In security-related content, a response that sounds convincing but is technically incorrect, an incorrect control, a distorted threat model, or an unsafe instruction marked as safe will be treated as an error. Rejected work is not paid, will be traced back to the annotator, must be redone on your own time, and may be penalized. You are expected to tell the difference between a truly correct security answer and one that merely appears credible. If a rubric point is genuinely unclear, you should flag the issue instead of guessing.

What you will do

• Review AI responses to security, privacy, and compliance prompts for technical accuracy and completeness.
• Tag security-relevant material using the provided taxonomy.
• Carry out pairwise and rubric-based preference assessments for cybersecurity outputs.
• Assign harm and safety labels to security content, including dual-use material where context determines risk.
• Classify prompt injection, jailbreaks, and other adversarial examples by attack type, and determine whether the model response was a safe refusal, a partial failure, or a complete bypass.
• Decide whether security advice in a model output is reliable or dangerous.
• Record short rationales that connect each decision back to the rubric.
• Stay aligned with the judgments of other security annotators.

Required profile

The right candidate should have practical or studied cybersecurity knowledge, including core security concepts, common attack vectors, web and network security, plus a working understanding of security governance and privacy. You should be able to separate technically correct content from plausible but wrong answers, and either already know or quickly learn how to apply prompt-injection and jailbreak taxonomy precisely. Strong reasoning, good English comprehension, and the discipline to follow a rubric consistently are essential.

Working setup and conditions

The role is remote and full-time, aligned to Lagos working hours, with coordination handled through the project tool and WhatsApp. Work is screen-proctored, camera-on, and done on a single device. You must have your own laptop and a reliable backup power source. ProtocolClear provides an internet and power stipend, but not equipment. Adversarial and harmful material must be handled only inside approved systems and never copied out.

Additional requirements

You must be comfortable with screening and proctoring requirements, willing to sign an NDA covering all client data, and prepared to handle harmful or adversarial content professionally. Unauthorized AI tools are not permitted while working.

Preferred background

Helpful experience includes security certifications such as Security+, CEH, or higher; work in a SOC, GRC, application security, or penetration testing environment; familiarity with NIST, ISO 27001, OWASP, and NIST AI risk material; prior RLHF, red-teaming, or model-evaluation experience; a coding or systems background; and experience writing security findings or rationales.

Compensation

The pay structure includes a monthly base amount, an internet and power stipend, per-task pay for accepted work, and performance bonuses for consistently strong quality. Estimated monthly earnings range from about ₦250,000 for light accepted output to about ₦1,500,000 for sustained full output at five-star quality, including base pay, stipend, and bonuses. This is a provisional range and will be revised once the client’s exact per-task terms are finalized. The first three months are a paid trial period with per-task rates plus stipend.

Other information

The role is being offered as a remote, full-time position from Lagos, Lagos State, Nigeria. The work is focused on security-domain annotation and evaluation for AI systems.