AI Model Protection Against Theft and Reverse Engineering

Skyld focuses on securing AI models throughout their entire lifecycle, with an emphasis on protecting models deployed on edge devices where exposure risks are high. The company provides an SDK that applies cryptographic controls and obfuscation techniques to prevent unauthorized access, cloning, or model misuse. These capabilities support intellectual property protection and regulatory compliance for organizations handling sensitive data.

Skyld also enables granular usage governance, including activation windows, geographic restrictions, and device-level permissions — features aligned with business models that rely on controlled distribution of proprietary AI models.

In addition to runtime safeguards, Skyld offers vulnerability assessments to identify weaknesses in AI models exposed to adversarial attacks or real-world threat conditions. The system incorporates algorithmic transformations designed to obscure internal model structures while retaining performance, allowing deployment in resource-constrained

environments without dedicated hardware. Through these combined protections, Skyld aims to help enterprises deploy AI systems securely across industries, reduce exposure to model theft and adversarial manipulation, and maintain operational integrity as cyber threats targeting AI systems continue to evolve.

In October 2025, Skyld announced that it had raised €1.5 million from French and Swiss investors: Auriga Cyber Ventures, BNP Développement, Bloomhaus, Sodero, and unnamed business angels.