White Circle

White Circle builds tools that control how AI systems behave, including guardrails, monitoring and red-teaming. Companies write rules for their AI in plain language, and the company's models check every input, output and tool call against those rules in real time. It also publishes AI safety and security research on how AI systems fail, including adversarial environments and behavioural evaluations.

The company sells to organisations that deploy AI systems for their customers or inside their own systems. The website offers a product demo request and a sales contact, and no pricing is published. The company describes its platform as processing tens of millions of requests per month.

White Circle was founded in 2025 and is a Public Benefit Corporation based in Paris and London. It states that it has raised $70 million from senior leaders and researchers at Hugging Face, OpenAI, Anthropic, DeepMind, Mistral and Datadog. Its careers page lists 18 open roles, mainly in research, go-to-market, design and operations, hired in Paris, London, New York, San Francisco, Los Angeles or remotely.

Product: Plain-language policy enforcement for prompts, model responses and agent tool calls, with red-teaming, monitoring and analytics.