When Brett Levenson left Apple in 2019 to lead business integrity at Facebook, the social media giant was dogged by Cambridge Analytica allegations. At the time, he thought he could easily fix Facebook’s content moderation problem with better technology.
He soon learned that the problem ran deeper than technology. Human reviewers were expected to memorize a 40-page policy document that was machine translated into their language, he said. They then had about 30 seconds per piece of flagged content to decide not only whether that content violated the rules, but what to do about it: block it, ban the user, limit the spread. According to Levenson, those quick calls were only “slightly better than 50% accuracy.”
“It was like tossing a coin, whether human reviewers could actually address the policies correctly, and that was several days after the loss was done anyway,” Levenson told TechCrunch.
This kind of delayed, reactive approach is not sustainable in a world of agile and well-funded adversarial actors. The rise of AI chatbots has further exacerbated the problem, as content moderation failures have resulted in a series of high-profile incidents, such as chatbots providing self-harm guidance to teenagers or providing AI-generated imagery bypassing security filters.
Levenson’s frustration led to the idea of ”policy as code” – a way to replace static policy documents with executable, updatable logic tightly coupled with enforcement. That insight led to the founding of Moonbounce, which announced it had raised $12 million in funding on Friday, TechCrunch has exclusively learned. The round was co-led by Amplify Partners and Stepstone Group.
Wherever content is generated, whether by user or by AI, Moonbounce works with companies to provide an additional security layer. The company has trained its own large language models to look at a customer’s policy documents, evaluate content at runtime, respond and take action in 300 milliseconds or less. Depending on the customer’s preference, that action may look like Moonbounce’s system is slowing down delivery while the content awaits human review later, or it may be blocking high-risk content at that time.
Today, Moonbounce operates in three main verticals: platforms dealing with user-generated content such as dating apps; AI companies are creating characters or companions; And AI image generator.
techcrunch event
San Francisco, CA
|
October 13-15, 2026
Levenson said Moonbounce is supporting more than 40 million daily reviews and serving more than 100 million daily active users on the platform. Customers include fellow AI startup Channel AI, image and video generation company Civity, and character roleplay platforms Dippy AI and Moscape.
“Security can actually be a product benefit,” Levenson told TechCrunch. “That’s never happened because it’s always been a thing that happens later, not something you can actually build into your product. And we see that our customers are finding really interesting and innovative ways to use our technology to make security a differentiator and part of their product story.”
Tinder’s head of trust and safety recently explained how the dating platform uses these types of LLM-powered services to reach a 10x improvement in identification accuracy.
“Content moderation has always been a problem that has plagued large online platforms, but now with LLM at the center of every application, the challenge is even more difficult,” Lenny Pruss, general partner at Amplify Partners, said in a statement. “We invested in Moonbounce because we envision a world where objective, real-time guardrails become the enabling backbone of every AI-mediated application.”
AI companies are facing increasing legal and reputational pressure as chatbots have been accused of driving teenagers and vulnerable users to suicide and image generators like XAI’s Grok have been used to create non-consensual nude imagery. Clearly, the safety guardrail is failing internally, and this is becoming a liability question. Levenson said AI companies are increasingly looking outside their walls for help strengthening security infrastructure.
“We’re like a third party sitting between the user and the chatbot, so our system isn’t loaded with context like chat is,” Levenson said. “The chatbot will have to remember, potentially, thousands of tokens that have come before… We are only concerned about enforcing the rules at runtime.”
Levenson runs the 12-person company with his former Apple colleague Ash Bhardwaj, who previously built large-scale cloud and AI infrastructure into the iPhone-maker’s core offerings. Their next focus is a capability called “iterative steering”, which was developed in response to cases such as the suicide of a 14-year-old Florida boy in 2024 who became obsessed with the character’s AI chatbot. Instead of outright denial when a harmful topic arises, the system will pause the conversation and redirect it, modifying prompts in real-time to push the chatbot toward a more proactively helpful response.
“We hope to be able to add to our toolkit of actions the ability to steer the chatbot in better directions, essentially taking the user’s cues and forcing the chatbot to be not just an empathetic listener, but a supportive listener in those situations,” Levenson said.
When asked if his exit strategy included an acquisition by a company like Meta, bringing his work on content moderation full circle, Levenson said he recognized how well Moonbounce would fit into his old employer’s stack, as well as his own fiduciary duties as a CEO.
He said, “My investors will kill me for saying this, but I would hate to see someone buy us out and then ban the technology.” “Like, ‘Okay, it’s ours now, and no one else can benefit from it.'”
