Test conversational AI with jailbreaks, prompt injections, and other adversarial scenarios while documenting vulnerabilities that automated tools may miss. This remote contractor role requires native English and Odia fluency and pays $16 to $22 per hour.
Generative AI & RLHF
$16–$22/hr
Compensation
30 countries
Eligibility
Entry
Experience
Jul 13, 2026
Posted
Open to applicants in
OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help contributors discover projects, build a professional profile, and apply to opportunities in minutes as they grow experience in this fast-moving field.
AI training is the human side of building modern artificial intelligence. People evaluate model responses, identify weaknesses, create structured examples, and provide feedback that helps AI systems become more reliable, useful, and safe.
Red teaming applies adversarial thinking to this process. By probing conversational models and agents with difficult or manipulative inputs, contributors can uncover risks that automated testing may not detect.
OpenTrain is seeking an AI Safety Red Teaming Expert for text-based testing of conversational AI models and agents. You will use adversarial inputs to explore how systems respond to challenging, manipulative, or unsafe scenarios, then produce structured human data to support safety improvements.
Assignments may involve sensitive topics such as bias, misinformation, or harmful behavior. Higher-sensitivity work is supported by clear guidance and wellness resources, and topics are communicated before exposure.
You will evaluate conversational AI across a broad range of adversarial scenarios. Your work will combine hands-on testing, careful annotation, vulnerability classification, and clear documentation that can be used by technical and non-technical teams.
Native fluency in both English and Odia is required. You should be able to assess whether AI responses are accurate, complete, and appropriate while noticing subtle errors, inconsistencies, and gaps.
The role also requires disciplined documentation and the ability to apply structured guidance across evaluation projects. Helpful experience includes adversarial machine learning, jailbreak datasets, prompt injection, RLHF or DPO attacks, model extraction, penetration testing, exploit development, reverse engineering, abuse analysis, misinformation probing, or conversational AI testing.
Backgrounds in psychology, acting, or unconventional creative writing may also support the adversarial thinking needed for this work.
This is remote contractor work available to candidates in the listed countries below. The role has a stated availability requirement of 20 or more hours per week and a default commitment of 40 hours per week.
AI safety red teaming is part of a rapidly growing industry where human judgment directly shapes how advanced AI behaves. Through OpenTrain, you can build a profile around your work, discover projects that match your capabilities, and develop a lasting career portfolio in AI training and data labeling.
Keep exploring
Use structured red team methods to uncover vulnerabilities in conversational AI systems and agents. This part-time contractor role pays $29-$45 per hour and requires native English and Portuguese fluency.
Posted Jul 30, 2026
Probe conversational AI for jailbreaks, prompt injections, bias, and misuse as an expert red teamer. Work remotely as a Dutch and English native speaker for $48 to $62 per hour, with a 20+ hour weekly commitment.
Posted Jul 30, 2026
Test conversational AI for jailbreaks, prompt injections, bias exploitation, and multi-turn manipulation. This expert contract role offers $48-$62 per hour and requires native English and Danish fluency.
Posted Jul 30, 2026
Browse related job pages
Locations
Languages
