Moonbounce raises $12M to automate AI content moderation

Moonbounce, an AI safety startup, announced on Friday that it has raised $12 million in a funding round co-led by Amplify Partners and StepStone Group. Founded by CEO Brett Levenson and Ash Bhardwaj, both former Apple employees, the 12-person company aims to replace slow, manual content moderation with ‘policy as code’—a method of turning static policy documents into executable, updatable logic. Levenson, who left Apple in 2019 to lead business integrity at Facebook, noted that traditional human moderation often relied on reviewers memorizing a 40-page policy document, giving them only about 30 seconds per piece of flagged content. This reactive approach resulted in decisions that were only slightly better than 50% accurate, often occurring days after the harm had already taken place.

To address this, Moonbounce provides an additional safety layer that evaluates content at runtime. The system is designed to evaluate content and provide a response in 300 milliseconds or less, allowing platforms to block high-risk content or slow down distribution for human review. The startup currently supports more than 40 million daily reviews and serves over 100 million daily active users. Its customer base includes AI companion startup Channel AI, image and video generation company Civitai, and character roleplay platforms Dippy AI and Moescape.

The funding arrives as AI companies face mounting legal and reputational pressure. Chatbots have been accused of pushing teenagers and vulnerable users toward suicide, and image generators like xAI’s Grok have been used to create nonconsensual nude imagery. Other safety failures include chatbots providing teens with self-harm guidance and AI-generated imagery evading safety filters. In response to cases like the 2024 suicide of a 14-year-old Florida boy who became obsessed with a Character AI chatbot, Moonbounce is developing ‘iterative steering’—the capability to intercept and redirect chatbot conversations in real time. Under this approach, the system would intercept conversations and redirect them, modifying prompts in real time.

Investors and founders emphasize that safety is shifting from a reactive burden to a product differentiator. Lenny Pruss, general partner at Amplify Partners, noted that content moderation has always been a problem that plagued large online platforms, but now with LLMs at the heart of every application, this challenge is even more daunting. Platforms using similar services have reported a 10x improvement in the accuracy of detections. “Safety can actually be a product benefit,” said Brett Levenson, CEO and founder of Moonbounce, who added that he would hate to see an acquisition restrict the technology.

Why it matters

Moonbounce is addressing the limitations of reactive content moderation by implementing ‘policy as code,’ a method of turning static policy documents into executable, updatable logic. This approach allows for the real-time enforcement of safety guardrails in AI applications, shifting safety from a delayed compliance measure to an active product feature.