Belgian Cybersecurity Firm Aikido Releases Open AI Model

Belgian cybersecurity company Aikido has launched an open-weight artificial intelligence model specifically designed for cybersecurity applications, addressing mounting concerns about code security and data privacy in enterprise environments.

The model, called Altar, represents a compressed and customized version of Z.AI's open-source GLM-5.3 model. According to details first reported by Reuters, Aikido will make Altar available for local deployment, allowing organizations to run AI-powered security tools on their own infrastructure rather than relying on cloud-based services.

Growing demand for on-premises security AI

The release comes as enterprises face increasing pressure to protect sensitive code and intellectual property. Criminals have begun leveraging AI to identify and exploit software vulnerabilities at scale, creating urgent demand for defensive tools that can match this threat without introducing new security risks.

Running security AI models locally addresses a critical concern: many companies are reluctant to send proprietary source code to external AI providers for analysis, even when those tools could identify vulnerabilities. By enabling on-premises deployment, Altar allows organizations to maintain complete control over their code while still benefiting from AI-powered security scanning.

Deployment and early adoption

Aikido plans to integrate Altar into its existing security solutions. Belgian bank Belfius is among the customers that will deploy the model, according to the company's announcement.

The firm has established itself as one of Europe's leading cybersecurity providers. In January, Aikido reached a $1 billion valuation, reflecting strong investor confidence in the European cybersecurity market.

Why it matters

The launch of Altar highlights a fundamental tension in enterprise AI adoption: organizations want the benefits of advanced AI tools but face legitimate concerns about data exposure. Open-weight models that can run locally offer a potential resolution, particularly in security contexts where code confidentiality is paramount. As AI-powered attacks become more sophisticated, the ability to deploy defensive AI without compromising proprietary information may become a competitive requirement for enterprise security vendors.

The details were first reported by Reuters on September 21.