AI Safety Experts: English & Urdu

What you'll do

Urdu poses a particular set of problems for AI models. It is written in the Perso-Arabic Nastaliq script, but a large share of everyday Urdu online (especially in Pakistan) is typed as Roman Urdu in Latin letters, and in speech it overlaps heavily with Hindi. A model can handle one of those forms safely and fail on another, or blur Urdu and Hindi in ways that carry political and cultural weight. Mercor wants native Urdu speakers to probe exactly those seams.

The ad's tasks:

• Red-team chat models and agents with jailbreaks, prompt injections, misuse cases, bias exploitation and multi-turn manipulation
• Annotate failures, classify vulnerabilities and flag systemic risks
• Work to taxonomies, benchmarks and playbooks
• Write reproducible reports, datasets and attack cases

As with the Indian-language variants it is grouped with, the Urdu profile is evaluation-led rather than security-led. Mercor asks for strong judgment about whether an AI response is accurate, complete and appropriate, attention to subtle errors, consistent use of guidelines and clear explanations. Prior red-teaming is welcome but not the headline requirement.

All work is text. Sensitive topics (bias, misinformation, harmful behaviour) are part of it; topics are communicated in advance, and higher-sensitivity projects are optional with guidelines and wellness resources.

Who fits

• Native fluency in Urdu and English
• Good judgment on AI answer quality, with reasons you can put into words
• Care with detail and consistency with quality standards
• Clear written communication
• Adaptability across tasks and clients

Nice to have: adversarial ML, cybersecurity, harassment and disinformation analysis, psychology, acting or writing. No degree or minimum years are stated.

What it pays

$16–22 per hour, Mercor's published band, the same as the Kannada, Malayalam, Tamil, Telugu, Marathi, Odia and Gujarati variants. It is the bottom band of this family; the European variants pay $48–62 for the same template.

Weekly payment via Stripe or Wise is stated in the ad. H-1B and STEM OPT candidates are not supported.

Worth knowing

Good:

• Evaluation-led profile, so careful native speakers without a security background can apply
• No residence requirement published, so speakers in Pakistan, India, the Gulf or the UK are all eligible on the ad's terms
• Weekly US-dollar payouts
• Sensitive projects are opt-in, topics disclosed first

Less good:

• No hires count was shown when we read the listing
• Lowest pay band in the family
• Red-team content can be distressing
• No weekly hours stated; projects can be extended, shortened or ended early
• Contractor terms; clients and models are not named

About this listing

A Mercor remote hourly contract, read from the Mercor site on 24 September 2026. Pay, requirements, payment terms and the visa exclusion come from the ad. Mercor states the work involves no confidential information from any employer, client or institution. See Mercor.