Pumpkin Intelligence Inc. (White Circle) raises $11M in Seed

Pumpkin Intelligence Inc., known as White Circle, has secured $11 million in seed funding to enhance its AI guardrail and monitoring technology. This funding underscores growing concerns about the security of AI models, which could be vulnerable to prompt injection and jailbreak attacks. Notable investors in this round include influential figures from the AI sector such as Romain Huet of OpenAI Group, Anthropic's Dirk Kingma, Mehdi Ghissassi, formerly of DeepMind, and Olivier Pomel, CEO of DataDog.

Denis Shilov, founder and CEO of White Circle, initiated the company following demonstrations that he could bypass existing safety protocols in major AI models with a single prompt. His research revealed vulnerabilities that could expose sensitive information or instruct AI models to perform harmful actions. White Circle offers a single API solution designed to monitor and secure AI model behavior by preventing inbound attacks and managing outputs. The system uses specialized AI models to dynamically detect and thwart potentially damaging activities based on user-defined policies.

The strategic focus of White Circle is to provide reliable guardrails as AI becomes increasingly integral in sectors such as healthcare, finance, and security. Shilov highlighted a potential risk: the rapid advancement of AI outpacing human ability to understand and guide its applications. The API developed by White Circle aims to preemptively identify issues like model drift and malicious user activity to enhance security and performance.

In the broader market context, White Circle’s initiative comes as AI models continue to reveal unpredictable behaviors, such as sycophantic responses in ChatGPT or jailbreak vulnerabilities in xAI Corp’s Grok model. Enterprises utilizing AI must increasingly rely on robust guardrails to handle sensitive data responsibly. The increasing gap between AI capabilities and security measures leaves room for specialized firms like White Circle to play a critical role in ensuring operational safety and integrity.

Looking ahead, White Circle aims to extend its monitoring capabilities globally, asserting compatibility across 150 languages and integrating with various AI products. The firm has also conducted extensive studies on AI biases across platforms like OpenAI and Google, emphasizing the ongoing need to address these challenges for secure and accurate AI deployment. As White Circle continues to expand, the company positions itself at the forefront of AI safety solutions, signaling a pivotal evolution in how AI integration and security are managed by businesses worldwide.