White Circle, a company specialising in enterprise AI control technology that tests, protects, monitors and enhances AI systems in real time, has raised $11 million in seed funding from a group of high-profile AI industry leaders.
The investment round included support from Romain Huet of OpenAI; Dirk Kingma, formerly with OpenAI and currently at Anthropic; Guillaume Lample of Mistral AI; Thomas Wolf of Hugging Face; Olivier Pomel of Datadog; François Chollet of Keras; Mehdi Ghissassi and Paige Bailey, both formerly associated with DeepMind; and David Cramer of Sentry. White Circle said the funding will support product development, team expansion across the US, UK and Europe, and growth of its international customer network.
The business was founded by engineer Denis Shilov, who drew industry attention in 2024 after revealing a single prompt capable of bypassing the safety protections of major AI models. The exploit uncovered instructions related to drugs and weapons and exposed harmful material that platforms such as ChatGPT and Claude were designed to restrict. The demonstration generated 1.4 million views and led to engagement with Anthropic, OpenAI and Hugging Face. Shilov later joined Anthropic’s bug bounty programme, an experience that contributed to the launch of White Circle in response to increasing demand for stronger AI oversight tools.
White Circle’s proprietary technology continuously analyses AI inputs and outputs while enforcing company-defined policies to identify harmful material, hallucinations, prompt injection attempts, model drift and abusive activity. The platform is designed to detect issues such as FinTech models exposing confidential customer information, AI agents responding to malicious instructions, users attempting to exploit systems, and declining model performance in real time. Organisations can implement customised enforcement measures, including rate limits and automated bans, while the platform improves through labelled feedback over time. The system supports 150 languages, operates through a single API, holds SOC 2 Type I and Type II certifications, and complies with HIPAA standards. Product design is overseen by head of design Elena Iumagulova.
White Circle describes itself as the only all-in-one enterprise platform dedicated to AI governance, arriving at a time when AI-powered products are expanding rapidly. This includes applications created through “vibe coding”, which enables users with limited technical expertise to deploy AI applications more easily, often ahead of established safety controls.
The company has also contributed to wider industry research into AI risk. In May 2025, White Circle introduced CircleGuardBench, a public benchmark designed to evaluate how AI moderation models perform in real-world environments and allow direct model comparisons. More recently, the company published its KillBench study, which involved more than one million experiments across 15 AI models developed by OpenAI, Google, Anthropic and xAI. The findings showed that when AI systems were tasked with making decisions involving human lives, every model displayed biases associated with nationality, religion, body type and even smartphone brand. The study further found that structured output methods commonly used in production AI systems significantly reduced refusal rates and intensified observed biases.
