ع
CyCraft announced that it had been identified as a sample provider in the Gartner Emerging Tech report titled AI Vendor Race: Reasoning Models Reshape Providers’ Approaches to AI Pricing, dated June 5, 2026. The assessment examined how the shift from training models to reasoning models with advanced logical deduction and self-reflection capabilities has driven up operational costs across the sector. According to the company’s statement, this inclusion positions CyCraft as the sole Taiwan-listed cybersecurity provider featured in the full list of vendors reviewed. Dr. Benson Wu, CEO and co-founder of CyCraft, described the recognition as evidence that AI security has become a primary focus for commercialization worldwide.
Wu stated, “We believe, being listed as a Sample Provider in the Gartner report demonstrates that the global market now views AI security as the next battlefield for AI commercialization.” He noted that CyCraft has built proprietary models applying AI reasoning capabilities across cybersecurity defense, AI governance and unmanned vehicle reconnaissance since its founding. The executive added that the company aims to convert this technical foundation into commercial expansion through targeted deployment and execution discipline. CyCraft launched AI Red Teaming and AI Guardrail solutions earlier this year that have already seen commercial adoption in semiconductor manufacturing and sovereign AI projects.
The company’s XCockpit platform incorporates AI reasoning models that process thousands of alerts in real time while generating mitigation recommendations to address alert fatigue. It consolidates metrics for external exposure, internal attack surface management and endpoint protection into unified scoring and actionable tickets. This approach turns traditionally qualitative defensive outcomes into measurable technical indicators, the statement explained. A Mordor Intelligence assessment of the AI cybersecurity sector projected sustained double-digit annual growth through the end of the decade driven by rising demand for such integrated platforms.
CyCraft has introduced XecART for agentic red teaming that conducts multi-round stress tests on large language models and generative AI systems to produce compliance reports aligned with international standards. Its companion product XecGuard functions as a real-time guardrail for LLMs and AI agents by intercepting anomalous inputs and outputs. Both solutions appear on the OWASP AI Security Map and support deployment in cloud or on-premises environments. The company said these tools tackle core challenges in agentic AI by addressing pre-deployment risk testing and runtime protection respectively.
The CyCraft research team presented findings on retrieval-augmented generation data poisoning at the International Conference on Machine Learning in 2026 using a self-developed Eyes-on-Me attack framework. The work demonstrated an increase in end-to-end attack success rates from 21.9 percent to 57.8 percent by targeting both retrieval and generation phases simultaneously. In 2025 the team contributed to NeurIPS with a unified supervision framework that enables self-supervised learning for safety robustness and instruction adherence without oracle labels. These academic contributions complement the firm’s product development in AI governance.
As reasoning models increase computational demands and force providers to adjust pricing expectations, enterprises face growing pressure to implement auditable AI systems particularly in financial services, government operations and critical infrastructure. CyCraft’s integrated suite links XecART for pre-launch testing, XecGuard for runtime monitoring and XCockpit for quantifiable risk metrics into a traceable control chain. The company, listed under stock code 7823 on the Taiwan exchange, maintains a track record serving government, finance and semiconductor clients with its AI-driven cybersecurity portfolio.
