Write a report
Platform AI rules, read November 18, 2026

AI for
bug bounty

AI for bug bounty helps with scope, recon notes, duplicate checks and reports. HackerOne, Bugcrowd, Intigriti and YesWeHack all allow it if you verify every finding yourself.

What each platform allows

Each card quotes the platform's own rule.

HackerOne

Encouraged

permits and encourages the responsible use of AI tools throughout the research workflow
  • AI for recon, discovery and reports
  • Validate every finding yourself
  • Fully autonomous hackbots
  • Disclosure of AI use required
Read the HackerOne rule

Code of Conduct. Updated for AI in May 2026.

Bugcrowd

Allowed

manually review and validate any vulnerability report you've created with the help of GenAI tools
  • GenAI to help write reports
  • Review every report by hand
  • Confidential data in AI tools
  • Disclosure of AI use required
Read the Bugcrowd rule

Code of Conduct. Modified December 1, 2025.

Intigriti

Encouraged

permitted and encouraged when used responsibly as part of a hacker's workflow
  • AI across your workflow
  • Disclose when and how you used AI
  • Verify the bug yourself
  • Unverified or fabricated reports
Read the Intigriti rule

Community Code of Conduct. Dated March 9, 2026.

YesWeHack

Controlled

maintain control over tools, scanners, payloads, and LLM results
  • LLMs under your control
  • Manual verification of findings
  • Reports from AI hypotheses
  • Disclosure of AI use required
Read the YesWeHack rule

Platform code of conduct. Dated March 16, 2026.

Full policy table

Where AI fits in a hunt

Four jobs, each inside the program's scope.

Before you test
Scope
Read the scope

Turn a program policy into a checklist

Get in-scope assets, exclusions, rate limits and AI rules as a list you check against the policy.

Read the guide
While you map
Recon
Sort your notes

Group recon output by asset and scope

Have a model group hostnames and notes you already collected, and flag anything outside scope.

Read the guide
Before you submit
Triage
Check for duplicates

Check exclusions, duplicates and severity

Compare a finding with the program's exclusions and disclosed reports, then rate it the way the platform does.

Read the guide
When you submit
Report
Write the report

Draft the report from your own notes

The model drafts in the platform's format. You check every step, claim and number.

Read the guide

Which models work

Ranked for security work, with prices per 1M tokens.

#ModelBest forInput / outputReads
1Kimi K3.1 ProBest for security work$3.50 / $8.00text, images, video
2GLM-5.3Reading code in scope$2.00 / $4.00text
3DeepSeek V4.1 Flash 1017Best value$0.24 / $0.90text, images
4GLM-5.3 FlashBulk notes$0.24 / $1.00text, images

See the full ranking or run a model locally.

Where to run them

Ranked on the facts that matter for program data.

ProviderModel for this jobUncensored Kimi K3.1 ProContextOn prompts, quoted
Silk ComputeKimi K3.1 ProYes1,048,576"Prompts and responses are never stored."
AbliterationAbliterated Large v2 (GLM-5.3)No1,000,000"never stored by default"
MorphDeepSeek V4.1 Flash UncensoredNo1,048,576"delete them within 30 days"

All seven providers compared. bugbountyai.com is published by the team behind Silk Compute, one of the providers ranked here.

Questions

Can I use AI for bug bounty?

Yes, on all four major platforms, if you verify every finding yourself. Intigriti also asks you to disclose how you used it (details).

What is the best AI for bug bounty?

We rank Kimi K3.1 Pro first for security work, since it reads screenshots and video and writes up to 943,718 tokens. DeepSeek V4.1 Flash 1017 is the best value at $0.24 in and $0.90 out per 1M tokens.

Can AI write my bug bounty report?

It can draft one from your notes, but every platform here penalizes reports nobody verified. Check every step and claim before you submit.

Can I paste program data into an AI tool?

Bugcrowd forbids GenAI use that discloses confidential information, so read the program rules first. Use a provider that does not store prompts, or a local model.