Aikido Security

Altar-1 is an open-weight AI model developed to support defensive cybersecurity tasks without requiring organizations to send internal context or source code outside their own environments. Built by Belgian application security company Aikido, it targets security teams that want to run vulnerability discovery and penetration testing workloads on infrastructure they control, with banks with data-residency obligations, healthcare providers handling sensitive patient information, and industrial organizations as clear early buyers. The launch matters because most frontier security tooling still requires shipping code and telemetry to a vendor cloud, and Altar-1 is a working alternative for teams that cannot. Under the hood, Aikido created Altar-1 from Z.AI's open-weight GLM-5.3 model, reducing its storage footprint from 1.51 TB to 328 GB. The company first applied AWQ INT4 quantization, reducing the model to 488.2 GB, then used expert pruning to remove components less relevant to security work, letting the model run on four Nvidia H200 GPUs while retaining 168 of the model's 256 routed experts . In Aikido's own evaluation, Altar-1 was tested against 32 known vulnerabilities spread across 30 repositories, with three runs for each case, and averaged 60.4% recall per run while finding 23 of the 32 vulnerabilities at least once. It powers locally deployed workloads in the Aikido Machine, the company's agent for automated security tasks. Aikido was founded on October 26, 2022 by Willem Delbare, Roeland Delrue, Felix Garriau and Madeline Lawrence, and is headquartered in Ghent, Belgium . The company raised a $17 million Series A in 2024 to expand its developer-facing security platform, and Altar-1 signals a shift from packaging third-party scanners toward training and shipping its own models for security workflows.