tech
April 3, 2026
The Facebook insider building content moderation for the AI era
Moonbounce has raised $12 million to grow its AI control engine that converts content moderation policies into consistent, predictable AI behavior.
TL;DR
- Brett Levenson left Facebook to found Moonbounce, aiming to solve content moderation problems exacerbated by AI.
- Moonbounce's "policy as code" converts complex policies into executable logic for real-time content evaluation and enforcement.
- The company raised $12 million, co-led by Amplify Partners and StepStone Group.
- Moonbounce serves platforms with user-generated content, AI character companies, and AI image generators.
- The technology evaluates content at runtime, responding in under 300 milliseconds, and can slow distribution or block high-risk content.
- A new feature, "iterative steering," aims to redirect conversations involving harmful topics toward more supportive responses.
- Levenson envisions safety as a product differentiator, citing Tinder's 10x improvement in detection accuracy using similar LLM services.
Continue reading the original article