In 2024, engineer Denis Shilov made headlines by demonstrating a universal jailbreak that bypassed safety filters on ChatGPT, Claude, and other leading models. Instead of exploiting the flaw, he founded White Circle, a Paris-based startup that has now secured $11 million in seed funding from a stellar roster of investors—including executives from OpenAI, Anthropic, DeepMind, Hugging Face, Mistral, Datadog, and Sentry.

From Vulnerability to Venture

Shilov’s jailbreak, which he presented at a security conference, caught the attention of Anthropic itself. Rather than patching the issue quietly, he turned his research into a business. White Circle’s platform offers a unified API for real-time monitoring of AI inputs and outputs, detecting hallucinations, prompt injections, harmful content, model drift, and abusive users. The startup claims to have already processed over 1 billion API requests and counts two of the world’s largest digital banks as customers, along with AI code generation startup Lovable.

All-Star Backing

The seed round’s backers include Romain Huet from OpenAI, Durk Kingma from Anthropic, Guillaume Lample from Mistral, Thomas Wolf from Hugging Face, and others from Datadog and Sentry. “This is not just a financial investment,” said Huet in a statement. “White Circle addresses a critical gap in AI observability that every enterprise deploying LLMs will need.” The involvement of rivals underscores a collective recognition that independent security solutions are essential—even as labs like OpenAI and Anthropic invest heavily in their own guardrails.

Platform Capabilities

White Circle’s platform provides real-time enforcement with support for over 150 languages. It also includes research contributions such as CircleGuardBench, a benchmark for moderation model performance, and KillBench, a study of biases across 1 million experiments. The company has achieved SOC 2 and HIPAA certifications, signaling readiness for regulated industries.

Competitive Landscape

White Circle enters a market with players like Arthur AI and WhyLabs, but Shilov argues that its real-time enforcement and language coverage give it an edge. “Most solutions focus on monitoring after the fact,” he explained in a press release. “We intercept risks before they reach users.” The platform’s single API approach simplifies integration, a key pain point for enterprises.

Market Opportunity

As generative AI moves from experimentation to production, governance has become a top concern. Regulations like the EU AI Act are forcing companies to implement robust safety measures. White Circle’s early traction with major banks validates demand. The $11 million will be used to expand its team across the US, UK, and Europe.

Looking back, the rise of AI safety startups mirrors the evolution of cybersecurity in the 2010s. In 2023, the OWASP Top 10 for LLM Applications highlighted prompt injection as a critical risk—a threat White Circle directly addresses. Similarly, the 2024 jailbreak incident echoed earlier vulnerabilities like the 2023 “DAN” jailbreak that exposed ChatGPT’s lack of robustness. These events have accelerated enterprise adoption of third-party security tools, just as the 2017 Equifax breach spurred investment in cloud security.

For comparative context, White Circle’s 150-language support and real-time enforcement set it apart from earlier AI monitoring tools that focused on batch analysis. Its investor lineup—drawing from rival labs—suggests that the industry is maturing toward collaborative safety standards, much like the formation of the OWASP foundation. As models grow more powerful, the attack surface expands, making infrastructure like White Circle’s not just a luxury but a necessity for responsible AI deployment.