AI training jobs for cybersecurity professionals
AI labs need people who can tell a working exploit from a convincing-sounding one. If you're a pentester, red teamer, SOC analyst, incident responder or AppSec engineer, several AI training platforms are paying for that judgment right now. Here's who's hiring, what each listing pays, and what they ask for. Every listing below was checked on October 30, 2026.
Last checked: October 30, 2026
The short answer
- Security-specific AI training roles are open right now at Mercor, Handshake AI, Outlier, DataAnnotation, Meridial and Alignerr. The work is mostly grading AI-written exploits, code and incident analysis, building test environments, and red-teaming model guardrails.
- Posted rates on October 30, 2026 ran from $15 to $250 an hour (USD), depending on the listing. These are listed rates and ceilings, not typical earnings. The $200–$250 rate is one Mercor listing for offensive-security researchers with public track records; most of the other listings post rates or ceilings between $60 and $150 an hour.
- Most listings we found are offensive-security roles: pentesting, vulnerability research and red-teaming. SOC and incident-response roles exist but are fewer.
- Lab bug bounties aren't jobs. They pay per accepted finding, not per hour, and each lab decides rewards at its discretion. OpenAI's Safety Bug Bounty and Anthropic's Model Safety Bug Bounty both publish reward ranges.
Cybersecurity AI training roles open now
| Platform | Role | Posted pay (USD) | Who it's for | Where |
|---|---|---|---|---|
| Mercor | Cybersecurity Research Expert – Offensive Security & Vulnerability Research (talent network) | $200–$250/hour | Top CTF players, CVE authors, recognized bug bounty researchers, security research labs | US [source] |
| Mercor | CVE Vulnerability Expert | $70–$90/hour | 3+ years in AppSec, pentesting or vulnerability research; Docker and Docker Compose | Remote, US [source] |
| Handshake AI | AI Red Teamer, Cybersecurity | $65–$125/hour | Background in offensive security, red teaming, malware analysis, threat intel or IR; can read and evaluate code used in security work; strong hands-on experience with multiple LLMs | Seattle or remote, US (not hiring in California) [source] |
| Handshake AI | Security Engineer (AI Fellowship) | Up to $80/hour, depending on the project | 3+ years in security engineering or offensive security; hands-on with tools like Burp Suite, Nmap, Metasploit or BloodHound | Remote [source] |
| Handshake AI | Adversarial Prompt Expert | $45, up to $65/hour, depending on prior adversarial-testing experience | Heavy LLM users skilled at testing model guardrails; red team or offensive security background a bonus | US-based [source] |
| Outlier | Offensive Cybersecurity Expert | Up to $70, $100 or $150/hour, depending on the recruiting page (Outlier says rates vary by expertise, skills assessment, location and project need) | 4+ years in pentesting, red teaming or offensive security; OSCP strongly preferred; Terraform or Ansible | US, UK, Canada, Australia, New Zealand [source] |
| DataAnnotation | Cybersecurity Expert, Penetration Tester, Application Security Engineer, Incident Responder | "Up to $125+/hr" (DataAnnotation's wording), depending on task difficulty and specialization | Hands-on experience in the role's field (security engineering, pentesting, AppSec, or incident response/DFIR/SOC); no degree required | Remote [source] |
| Meridial | CVE & Application Security AI Task Auditor | $60/hour (final rate depends on experience, expertise and location) | Professional AppSec experience: vulnerability assessment, secure coding, threat modeling | Worldwide, remote [source] |
| Alignerr | AI Red Team Analyst | $15–$75/hour | Solid grasp of threat modeling, pentesting or vulnerability research, plus hands-on LLM experience; infosec background a bonus but not required | Remote; 10–40 hours/week (no posting date shown; open when checked 10/30/2026) [source] |
Pay is what each listing showed on October 30, 2026. Listings open, close and change rate often, so check the listing before you count on a number. "Up to" figures are ceilings, not what most people earn.
What the work actually looks like
- Grading AI-written attack work. Mercor's CVE role reviews whether CVE reproductions are faithful, fixes are sound, and Docker lab environments really recreate the exploitable condition. DataAnnotation's Penetration Tester role asks you to catch exploit chains that don't hold up and write the correct analysis when the model falls short. [source][source]
- Red-teaming the guardrails. Handshake AI's cybersecurity red teamers run authorized tests of whether a model's safety guardrails hold up against misuse, then score any failures against structured harm and severity rubrics so the lab can fix them. [source]
- Building test environments. Outlier's offensive-security project involves designing CTF-style attack environments across network, web app, cloud and Active Directory, deployed with Terraform and Ansible. [source]
- Defensive work. DataAnnotation's Incident Responder role grades how models handle detection, triage, forensics and containment. Common backgrounds it lists include DFIR analysts and SOC engineers. [source]
AI lab bug bounties (not jobs: paid per accepted finding)
These are bounty programs, not jobs. You join or apply, test within each program's rules, and get paid only for findings the program accepts. Reward amounts are at each lab's discretion.
| Program | What it pays for | Posted rewards (USD) | Source |
|---|---|---|---|
| OpenAI Safety Bug Bounty (on Bugcrowd) | AI abuse and safety issues: prompt injection and data exfiltration in agentic products, leaks of proprietary information, account and platform integrity bypasses. Jailbreaks are out of scope for this program. | $250–$500 (P4) up to $5,500–$7,500 (P1); final decisions and amounts at OpenAI's discretion | [source][source] |
| Anthropic Model Safety Bug Bounty (on HackerOne) | Authorized testing of whether its Constitutional Classifiers safeguards can be bypassed across many prompts (universal jailbreaks), focused on biological-threat safeguards. Application required; participants sign an NDA. | Up to $35,000 per novel, universal jailbreak, on a sliding scale at Anthropic's discretion | [source] |
Anthropic also runs a separate Cyber Jailbreak disclosure program on HackerOne for reports that its cyber safeguards can be bypassed. HackerOne lists it as a program that doesn't offer bounties. [source]
How to qualify
- Show offensive credentials if you have them. OSCP comes up again and again: Outlier says it's "strongly preferred" and Mercor's CVE role lists OSCP, GPEN or GWAPT as preferred. [source][source]
- Public work counts at the top end. Mercor's $200–$250/hour research role looks for several of: a top-ranked CTF team, CVEs you discovered, accepted bug bounty findings at major programs, or published security research. [source]
- Experience is the main filter. Listings ask for 3+ years (Mercor CVE, Handshake AI Security Engineer) or 4+ years (Outlier). DataAnnotation says no degree is required; Outlier's pages list a master's degree. [source][source][source]
- Use LLMs a lot. Handshake AI's red-team role asks for strong hands-on experience with multiple models, and its Adversarial Prompt Expert role centers on testing model guardrails. [source][source]
- Expect an assessment. Most platforms screen with a test or AI interview before you see paid work. Tips: How to get accepted on AI training platforms.
Tips
- Apply to more than one. Projects start and pause with client demand, so one platform may not have work for you right away.
- Check location first. Most of the top-paying security roles are US-only, and Handshake AI's red-team role excludes California. Outlier's project is open in five English-speaking countries; Meridial's lists worldwide.
- Write like your reports. Most listings stress clear written explanations. Your reasoning is what gets graded, so write the way you'd write a pentest finding: what's wrong, why, and the fix.
- Read the payout terms. Mercor pays weekly through Stripe or Wise; Outlier through PayPal or Airtm. See Mercor vs Outlier for details. [source][source]
- Expect sensitive material. Red-team work means testing models against misuse within each program's rules, and some of what you review can be harmful. Handshake AI's listing asks for "strong ethical judgment." [source]
FAQ
Do I need AI or machine learning experience?
Usually not. Most listings want security expertise plus comfort using chatbots. Alignerr does ask for hands-on LLM or prompt-engineering experience, and Handshake AI's red-team role wants heavy LLM use. [source]
Are these full-time jobs?
No. Mercor, Outlier, Meridial, Alignerr and DataAnnotation describe the work as independent contractor or freelance. Handshake AI's red-team role is listed as a contract with part-time availability. [source][source]
Is there work for SOC analysts, not just pentesters?
Yes, but fewer roles. DataAnnotation's Incident Responder role lists SOC engineers and DFIR analysts as common backgrounds. A Mercor paid-interview listing for SOC and incident-response practitioners had already closed when we checked. [source]
Can I do this outside the US?
Some of it. Outlier's offensive-security project accepts applicants from the US, UK, Canada, Australia and New Zealand. Meridial's AppSec auditor role is listed worldwide, with the rate set partly by location. Check our countries page for each platform. [source][source]
Is a bug bounty better than an hourly role?
They're different. Hourly roles pay for your time. Bounties pay only for accepted findings, but a single top-severity report can be worth thousands. OpenAI's Safety Bug Bounty posts up to $7,500 for its highest priority level. [source]
Try it
See Mercor (affiliate link: we may earn a fee if you sign up, at no cost to you). Or compare every platform's published pay on our pay page and platforms page.
Pay and listings checked October 30, 2026. Rates and openings change often; check each listing before you apply.