Find your job

Anthropic direct contractor program what it pays and how to enter.

Anthropic runs frontier AI training programs with rates above mainstream platforms. Here's what we know about the program, who qualifies, and the realistic entry paths.

Anthropic operates the most respected direct AI contractor program in 2026. Pay sits above what mainstream platforms offer. Entry is hard. Most contractors don't realize it exists. Here's the picture, drawn from public information and contractor reports.

What Anthropic's contractor program looks like

Anthropic engages contractors directly for: red teaming, constitutional AI evaluation, frontier reasoning evaluation, multi-step agent assessment, and specialty domain RLHF (medical, legal, scientific). Contracts are project-based, typically 4–12 weeks, with deeper engagement than standard gig platforms.

What it pays

  • Generalist contractor: $80–$130/hr
  • Senior with relevant background: $130–$200/hr
  • Specialty (medical, legal, security): $150–$280/hr
  • Bounty programs (responsible disclosure): $1,000–$25,000+ per finding

Among the highest contractor rates in AI training. The trade-off is volume — Anthropic engages a small number of contractors with high pay rather than a large pool.

Who qualifies

Anthropic sources contractors heavily through three channels:

  1. Existing senior contractors on partner platforms (Outlier, Mercor). Top-decile performers in specialty tracks get noticed.
  2. AI safety community engagement. Active contributors to LessWrong, Alignment Forum, AI safety research are visible to recruiters.
  3. Direct application via Anthropic's careers page. Lower hit rate than the above but possible.

What works less well: cold LinkedIn messages, generic application emails, traditional recruiter introductions.

Anthropic monthly income$150/hr × 12 hrs/wk = $7,500/month with bursty schedule.
Open calculator →

The credentials that move the needle

  • Demonstrated AI safety thinking. Public writing on alignment, evaluation, or interpretability.
  • Senior software engineering background. Particularly distributed systems or compilers experience.
  • Verified domain credentials. MD/JD/PhD in fields where Anthropic actively trains models.
  • Security research track record. CVEs, bug bounties, security publications.

Generic "I'm a strong engineer" doesn't score. Specific demonstrated work in one of these areas does.

The application paths that work

Path 1: tier up via mainstream platform

Reach senior tier on Mercor or Outlier in a specialty track (constitutional AI, agent eval, medical RLHF). Maintain 0.93+ scores for 90+ days. The platforms surface top-decile performers to Anthropic's contractor pipeline.

Realistic timeline: 6–12 months from starting on Mercor.

Path 2: AI safety community engagement

Publish substantive work on alignment, evaluation, or interpretability. Engage thoughtfully with active researchers. Anthropic recruiters actively monitor these communities.

Realistic timeline: 6–18 months of consistent contribution.

Path 3: direct application

Apply to specific contractor roles posted on careers.anthropic.com. The hit rate is low but the path is open.

Realistic timeline: 1–4 weeks for response if you fit the role spec exactly.

What working with Anthropic feels like

Reports from contractors who've worked direct programs suggest:

  • Higher quality bar than mainstream platforms — specific feedback, fewer repeated tasks.
  • Less stable income than Outlier/Mercor. Project gaps are real.
  • Stricter NDAs than mainstream platforms. Public discussion of work is restricted.
  • Career value is meaningful — the affiliation opens doors at AI/ML companies generally.

What Anthropic specifically isn't looking for

  • Generic crowdworkers expecting to "scale up" from labeling.
  • Contractors who can't articulate a specific safety or evaluation interest.
  • Applicants who treat the work as a side hustle to maximize hours.

Bottom line

Anthropic's direct contractor program is real, well-paid, and hard to enter. The two viable paths are: (1) reach top-decile performance on Mercor/Outlier in a specialty track, or (2) build a public AI safety / evaluation track record. Direct application is possible but lower-yield. If you're drawn to frontier evaluation work and have the right background, this is the highest-leverage AI contractor work available.

Find AI training contractsAll open roles · 9 platforms · filter by rate and hours.
Find your job

Frequently asked questions

How much does Anthropic pay contractors?
Generalist contractors earn $80–$130/hr. Senior contractors with specialty backgrounds reach $130–$200/hr. Domain specialists in medical, legal, or security can reach $150–$280/hr. Bounty programs pay $1,000–$25,000 per significant finding.
How do I apply to Anthropic's contractor program?
Three paths: (1) reach top-decile performance on Mercor/Outlier in a specialty track, (2) build public AI safety or evaluation work that's visible to recruiters, (3) direct application via Anthropic's careers page. The first two yield more interviews.
Is Anthropic's program more competitive than Mercor?
Significantly. Anthropic engages a small number of contractors at high pay; Mercor engages thousands. The bar at Anthropic is higher and the entry process is less standardized.
Do you need AI safety background to work with Anthropic?
Not strictly, but it helps significantly. Public engagement with AI safety / alignment research, even at amateur level, is a strong signal. Pure software engineering credentials alone are usually insufficient.