Just launched the MVP of a bug bounty platform for LLMs
I'm a UX researcher working on LLM products. I noticed quickly that while LLMs can bring value to certain products, their vulnerabilities make them huge liabilities for brand reputation and user experience. If you're developing a product with a user-facing LLM input field, there's a non-zero chance that it'll behave in ways you absolutely don't want it to.
An LLM is only as safe as the creativity of the people who tested it, so I hacked together a bug bounty platform for LLM products: promptbounty.io
PromptBounty connects companies and agencies with a diverse set of people to find out how the LLM can break. The fun part about PromptBounty is that, unlike standard bug bounties, you don't need to know to code to participate.
It may not be possible to build a fully secure LLM product, but it's possible to be mindful of the risk and design accordingly.
I'm looking for:
- People with LLM products who would like to have them tested
- People interested in being testers
- General feedback and connections!
Thank you!
