Moonbounce Raises $12M to Turn AI Policy Into Code
Content moderation has always been a game of whack-a-mole. Write a policy doc, translate it into rules, hope the rules catch what the policy intended. Moonbounce just raised $12M co-led by Amplify Partners and StepStone Group to replace that entire pipeline with what they call "policy as code" β a system that compiles human-readable policy documents into executable, updatable logic that enforces AI behavior in real time.
The founding team is purpose-built for this. Brett Levenson ran Meta's Integrity unit, the team responsible for content moderation across Facebook and Instagram. Co-founder and CTO Ash Bhardwaj built large-scale cloud and AI infrastructure at Apple. They've seen the problem from both the policy side and the engineering side.
The traction numbers are already serious. Moonbounce is processing over 40 million daily reviews and serving more than 100 million daily active users. Customers include AI companion startup Channel AI, image and video generation platform Civitai, and character roleplay platforms Dippy AI and Moescape. These aren't hypothetical customers β they're live, high-volume AI products that need policy enforcement at scale.
As agents become more autonomous, the gap between "what we want agents to do" and "what agents actually do" becomes the biggest risk surface. Moonbounce is betting that gap is a policy compilation problem, not a model alignment problem. Different thesis, same urgency.
https://techcrunch.com/2026/04/03/moonbounce-fundraise-content-moderation-for-the-ai-era/
← Back to all articles
The founding team is purpose-built for this. Brett Levenson ran Meta's Integrity unit, the team responsible for content moderation across Facebook and Instagram. Co-founder and CTO Ash Bhardwaj built large-scale cloud and AI infrastructure at Apple. They've seen the problem from both the policy side and the engineering side.
The traction numbers are already serious. Moonbounce is processing over 40 million daily reviews and serving more than 100 million daily active users. Customers include AI companion startup Channel AI, image and video generation platform Civitai, and character roleplay platforms Dippy AI and Moescape. These aren't hypothetical customers β they're live, high-volume AI products that need policy enforcement at scale.
As agents become more autonomous, the gap between "what we want agents to do" and "what agents actually do" becomes the biggest risk surface. Moonbounce is betting that gap is a policy compilation problem, not a model alignment problem. Different thesis, same urgency.
https://techcrunch.com/2026/04/03/moonbounce-fundraise-content-moderation-for-the-ai-era/
Comments