AI Safety Experts: English & Thai

What you'll do

Thai is a hard case for AI safety systems. It is written without spaces between words, uses its own script, carries a layered register system and deep cultural sensitivities, and is thinly represented next to English in model training. That makes it fertile ground for red-teaming, and Mercor's project wants native Thai speakers to probe it.

The brief:

• Attack conversational models and agents with jailbreaks, prompt injections, misuse cases, bias exploitation and gradual multi-turn manipulation
• Turn outcomes into labelled data: annotate failures, classify vulnerabilities and flag systemic risks
• Follow the project's taxonomies, benchmarks and playbooks so results are consistent
• Write reports, datasets and attack cases that the client can reproduce

A native Thai reviewer can recognise when an answer is offensive in a Thai social or political context, when a model mishandles honorifics or register in a way that changes meaning, and when switching script or mixing in English loosens a safeguard.

All tasks are text. Expect content on sensitive subjects (bias, misinformation, harmful behaviour); topics are announced first, and higher-sensitivity work is optional with guidelines and wellness resources.

Who fits

• Native Thai and native-level English, both required
• Some prior red-teaming: AI adversarial testing, cybersecurity, or socio-technical probing
• Disciplined, framework-driven testing habits
• Clear written explanations for technical and non-technical readers
• Comfortable across changing projects and customers

The ad also welcomes adversarial ML, pentesting and reverse engineering, harassment or disinformation analysis, and creative backgrounds in psychology, acting or writing. No degree or experience years are stated.

What it pays

$24–35 per hour, Mercor's published figure. That is a clear step below the $48–62 paid for the Nordic and Dutch versions of this same listing, and above the $17–25 offered for Indonesian, Vietnamese and Malay. The template and the experience bar are the same across all of them, so the difference reflects Mercor's market rate by language rather than harder work.

Weekly payment via Stripe or Wise is stated. H-1B and STEM OPT candidates cannot be supported.

Worth knowing

Good:

• High activity: 205 hires on this listing this month
• Competitive for Thai-language remote work paid in US dollars
• No residence requirement published, so Thai speakers abroad can apply
• Optional participation in the most sensitive projects

Less good:

• Same red-teaming bar as the top-band variants for roughly half the rate
• Disturbing material is part of the job, with safeguards
• No weekly hours stated; projects can be extended, shortened or ended early
• Independent contractor, with tax and benefits on you
• Clients and target models are unnamed

About this listing

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