CyCraft, a recognized leader in AI-driven cybersecurity, has announced the global release of XecGuard, a pioneering security solution designed to protect Large Language Models (LLMs) from emerging threats. This plug-and-play LoRA security module represents a major leap in AI defense, built specifically to counter risks like prompt injection, jailbreak attempts, and prompt extraction—vulnerabilities that have become increasingly common as enterprises rapidly integrate AI into critical operations.
Despite the growing reliance on LLMs, security has often been treated as an afterthought. CyCraft aims to change that narrative, asserting that protection must be a foundational component of AI systems. Without it, organizations risk exposing sensitive information, destabilizing services, and losing user trust. With XecGuard, CyCraft shifts the focus from reactive defense to proactive protection, embedding cybersecurity directly into the AI development lifecycle.
As businesses across industries—from government to finance to manufacturing—adopt AI at scale, the associated attack surface expands significantly. CyCraft draws on years of experience defending high-value sectors to build a solution tailored to modern AI environments. According to CEO Benson Wu, “AI security is no longer optional—it’s a standard requirement.” CTO Jeremy Chiu added, “We used to use AI to solve cybersecurity problems. Now, we’re using AI to defend itself.”
XecGuard is engineered for quick deployment, requiring no changes to existing LLM architecture. It offers compatibility with both cloud platforms (like AWS and Cloudflare Workers AI) and on-premises NVIDIA-based infrastructure, and is also available as a SaaS or an OpenAI-compatible firewall. The system has been tested extensively on open-source models such as Llama 3B, Qwen3 4B, Gemma3 4B, and DeepSeek 8B—demonstrating strong performance in both security enhancement and processing efficiency. In a partnership with NVIDIA-affiliated APMIC, XecGuard was integrated into the F1 open-source model, yielding a 17.3% average increase in defense scores during LLM Red Teaming assessments, and up to 30.1% in specific attack vectors.
With the release of XecGuard and its broader Safety LLM services, CyCraft is helping enterprises secure their AI deployments from day one. The goal is not only to protect sensitive data, but also to create a stable, trustworthy AI ecosystem where innovation can thrive. Backed by recognition from global institutions like Gartner and IDC, and with presentations at events such as Black Hat, DEFCON, and EMNLP, CyCraft continues to lead the charge in building secure, AI-powered futures.
