Break AI, get paid
Get an AI model to break its own rules, show how you did it, and win a prize or a bounty. Each program below lists its payout.
Contests and bug bounties
- Open nowBug bounty· 0DIN by Mozilla
0DIN GenAI Bug Bounty
Find jailbreaks, hidden-prompt leaks and other ways past the security limits of AI models and apps.
- $100 to $15,000 per finding
- Ongoing
Go to 0din.aiChecked Nov 14, 2026
- Open nowBug bounty· Google
AI Vulnerability Reward Program
Find security and abuse bugs in Google's AI products, such as the AI taking actions it should not, or leaking data.
- Up to $20,000 per report
- Ongoing
- Participation in the program is restricted to individuals aged 18 and older.
Go to bughunters.google.comChecked Nov 14, 2026
- Open nowBug bounty· Microsoft
Copilot Bounty Program
Find vulnerabilities in the Copilot AI experiences on copilot.microsoft.com and copilot.ai.
- $100 to $30,000 per finding
- Ongoing
Go to microsoft.comChecked Nov 14, 2026
- Open nowBug bounty· OpenAI
OpenAI Bug Bounty
Report security bugs in ChatGPT, the API and OpenAI's other services.
- $250 to $30,000 per finding
- Running since 2023-04-11
- Open to anyone with a Bugcrowd account
Go to bugcrowd.comChecked Nov 14, 2026
- Open nowBug bounty· OpenAI
Safety Bug Bounty
Report abuse and safety risks in OpenAI's agents, tools and platform.
- $250 to $7,500 per finding
- Running since 2025-07-29
- Open to anyone with a Bugcrowd account
Go to bugcrowd.comChecked Nov 14, 2026
- Contest· HackAPrompt
HackAPrompt
Prompt-hacking competition with four tracks.
- Over $100,000 in prizes given away so far
Go to hackaprompt.comChecked Nov 14, 2026
- Bug bounty· Anthropic
Model Safety Bug Bounty
Find one jailbreak that works on many prompts and gets past Anthropic's safety defenses.
- Up to $15,000 per universal jailbreak
- Invite only, through HackerOne
Go to hackerone.comChecked Nov 14, 2026 · Source: anthropic.com
- EndedContest· Gray Swan
Gray Swan Arena
Timed red-team challenges against AI models and agents. Latest: Gray Swan Launches Redesigned Arena Platform With $300K Challenge to Find the Next Generation of AI Red Teamers
- $300,000 in prizes sponsored by Anthropic, Meta, Google DeepMind, and Reflection AI
- October 15, 2026 to November 5, 2026
- Participants must be 18 years of age or older.
Go to app.grayswan.aiChecked Nov 14, 2026 · Source: grayswan.ai
Paid red-team and AI safety roles
Prefer an hourly rate? These AI training roles pay you to test models on a set project.
- AI Jailbreak & Prompt-Injection Security ExpertApply on micro1
micro1 · $50-$90/hr · Paid partnership
- Behavioral Health Expert, AI Safety and Model Evaluation - review AI outputs in your specialtyApply on Mercor
Mercor · $45-$70/hr · Paid partnership
- Bilingual Chinese Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $48-$52/hr · Paid partnership
- Bilingual Danish Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $48-$52/hr · Paid partnership
- Bilingual French Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $48-$52/hr · Paid partnership
- Bilingual Hindi Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $18-$22/hr · Paid partnership
- Bilingual Polish Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $40-$44/hr · Paid partnership
- Bilingual Portuguese Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $40-$44/hr · Paid partnership
- Bilingual Spanish Generalist Expert — AI Safety - review AI outputs in your specialtyApply on Mercor
Mercor · $43-$47/hr · Paid partnership
- LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness)Apply on Mercor
Mercor · $100-$120/hr · Paid partnership
- Responsible AI / Security Specialist - review AI outputs in your specialtyApply on Mercor
Mercor · $90-$140/hr · Paid partnership
- AI Safety & Evaluation SpecialistApply on micro1
micro1 · $30-$50/hr · Paid partnership
Common questions
- What is AI red-teaming?
- Testing an AI model the way an attacker would: getting it to ignore its safety rules, leak hidden data or do things it should refuse. AI labs pay people who find these breaks first. Contests share a prize pool. Bug bounties pay for each finding.
- Contest or bounty: what is the difference?
- A contest runs for set dates and shares a prize pool among the best attacks. A bounty pays for each new problem you report, based on how serious it is.
- Where do the prizes and payouts come from?
- We read each program's own page every day and show its payout. The Checked date on each card is when we last read it. Each program's own rules say who can join and what it pays.
- Can I get paid by the hour instead?
- Yes. The paid red-team roles below are hourly AI training jobs. You test models for a platform on a set project.