HackerOne AI Red Teaming
Time-boxed adversarial testing of AI models, agents and apps by vetted outside researchers.
Offered by HackerOneRed teaming and evaluation
About HackerOne AI Red Teaming
H1 AI Red Teaming is a HackerOne service that applies adversarial testing across prompts, models, APIs, and integrations.
The page says that AI red teaming validates that customers can only access authorized data and prevents cross-account leakage.
Drafted from hackerone.com · not yet confirmed by HackerOne
What they do
HackerOne runs AI red teaming as fixed engagements of 15 or 30 days, which it says can be scheduled in about a week. Testing is human-led by AI-focused researchers from its community, working alongside adversarial agents, and probes prompts, retrieval pipelines and agent workflows for jailbreaks, prompt injection, data leakage, tool misuse and unsafe or biased outputs.
It says more than 750 AI-focused researchers take part and that it has tested more than 1,700 AI assets. Named customers include frontier AI labs and technology companies, and it cites a challenge for Anthropic that produced more than 300,000 interactions. Findings are mapped to frameworks it lists, including the OWASP LLM Top 10, MITRE ATLAS, NIST AI RMF and the EU AI Act.
How they work
- How you engage them
- Project
Reviews
Be the first to review it
AnonymousKnow a team weighing HackerOne AI Red Teaming? Send them this page.