Skyld
Protects AI models from theft, extraction, and adversarial attacks.

Skyld
Protects AI models from theft, extraction, and adversarial attacks.
The Entire Cybersecurity Market, One Prompt Away
Connect your AI assistant to 10,000+ tools and 5,000+ vendors. Ask anything about the cybersecurity market.
Skyld Description
Skyld is a deeptech startup founded in July 2023 by Marie Paindavoine, a cryptography PhD with over ten years of cybersecurity experience. The company develops security solutions specifically designed to protect artificial intelligence models from sophisticated attacks including theft, reverse engineering, model extraction, and adversarial examples. The company addresses the security challenges that arise when AI models are deployed in unsecured environments across critical sectors such as energy, healthcare, defense, mobility, and industry. Skyld's approach focuses on ensuring the integrity, confidentiality, and monetization control of AI models. The founder's background includes expertise in homomorphic encryption, an advanced technology that enables computations on encrypted data, which is relevant to Privacy Preserving Machine Learning. Skyld operates on the principle of "AI security by design," recognizing that traditional cybersecurity protections are insufficient against attacks targeting AI systems. The company invests 85% of its budget in research and development. Skyld has been supported by Inria Startup Studio and accelerated by Berkeley SkyDeck. The company raised €1.5 million in its first funding round in October 2025, backed by French and Swiss investors including Auriga Partners, Axeleo Capital, and Verve Ventures. The startup has received recognition in the European cybersecurity ecosystem, including winning the Access-2-Finance Award at European Cyber Week 2023 and reaching the finals of the ECSO Startup Award 2024 among the 12 most promising cybersecurity startups in Europe.
POPULAR
TRENDING CATEGORIES
Stay Updated with Mandos Brief
Get strategic cybersecurity insights in your inbox