White Circle, a Paris company that builds a control layer for AI systems in production, announced an $11m seed round on 12 May 2026. The money comes from angel investors rather than a lead venture fund, and Fortune named four of them: Romain Huet, head of developer experience at OpenAI; Durk Kingma, an OpenAI co-founder now at Anthropic; Guillaume Lample, co-founder and chief scientist at Mistral; and Thomas Wolf, co-founder and chief science officer at Hugging Face.
The Next Web added Olivier Pomel of Datadog, François Chollet, creator of Keras, Mehdi Ghissassi, formerly of DeepMind, Paige Bailey of DeepMind and David Cramer of Sentry. SecurityWeek described the investors only as multiple angels. No valuation was disclosed.
“AI is moving faster than our ability to guide it”Denis Shilov, founder, White Circle
What White Circle does
The company's software sits between a company's users and its AI models. Fortune said it checks inputs and outputs in real time against the company's own policies, and can flag or block prohibited requests, hallucinations, data leaks and unauthorised actions. SecurityWeek said it also detects prompt injection, model drift and abusive users, and supports 150 languages.
The Next Web reported that customers can set enforcement actions such as rate limits and bans, that the company is SOC 2 Type I and Type II certified and HIPAA-compliant, and that it has served more than one billion API requests. The named customer is Lovable, the vibe-coding startup, along with several fintech and legal companies in Fortune's account and two of the world's largest digital banks, unnamed, in The Next Web's.
The founder
White Circle was founded in 2025 by Denis Shilov. The Next Web said he went viral in 2024 when a single prompt bypassed the safety filters of every major AI model; the post reached 1.4 million views and drew contact from Anthropic, OpenAI and Hugging Face, and he later joined Anthropic's bug-bounty programme.
The company has also published research. CircleGuardBench, from May 2025, tests how AI moderation models perform under real-world conditions. KillBench ran more than one million experiments across 15 models from OpenAI, Google, Anthropic and xAI and found preferences tied to nationality, religion, body type and phone brand when models made decisions about humans, The Next Web said.
Shilov told Fortune: "Jailbreaks are just one part of the problem."
What the money is for
White Circle has 20 people across London, France, Amsterdam and elsewhere in Europe, almost all of them engineers, according to Fortune. It said the funding will go to expanding the team, speeding product development and growing the customer base in the US, UK and Europe.
Shilov said: "AI is moving faster than our ability to guide it." He added that companies already trust AI with decisions that touch millions of people, from hiring to healthcare, finance and security.
Ophelia Cai, a partner at Tiny VC who is not listed as an investor, said Shilov and the team have an unusual combination of deep technical credibility and a clear commercial instinct.