#Japan #Tokyo #AI Security #CyCraft #CODE BLUE 2025

CyCraft at CODE BLUE 2025: Pioneering AI Security Solutions

In the rapidly evolving landscape of artificial intelligence, ensuring security has become paramount. This year, CyCraft proudly participated in CODE BLUE 2025, an esteemed conference dedicated to AI security advancements. They presented significant findings, particularly in improving safety measures for large language model (LLM) applications.

Highlights from the Session

CyCraft's presence at this event not only showcased their expertise but also highlighted their commitment to addressing contemporary security challenges posed by AI technologies. One of the key focus areas was the introduction of their cutting-edge solutions, such as the Exposure Attack Surface Management (EASM) and innovative guardrails and evaluation tools designed specifically for generative AI.

Keynote Overview

During the conference, CyCraft's Japan Architect Consultant, Renata Chang, delivered an insightful session titled Practical AI Security: Defending Against Injection Attacks with Guardrails. As LLMs become increasingly prevalent, new attack vectors such as Prompt Injection, Prompt Extraction, and Jailbreak techniques have surged, leaving traditional security measures insufficient to counter these threats.

Two Pillars for Defending AI Systems

Renata's session underscored two critical pillars essential for fortifying AI systems against evolving attacks:

1. Adversarial Red Team Testing: This approach involves revealing hidden vulnerabilities in LLMs through extensive injection scenario simulations. By identifying weak points, organizations can preemptively address these flaws before they are exploited by malicious actors.

2. Guardrail Defense: Participants learned about implementing verifiable multi-layer controls. This method aims to deter harmful prompts and ensure that AI models operate safely and predictably. Effective guardrails can enhance trust in AI applications, making them safer for broader implementation.

Utilizing Real-World Validation

An important aspect of the presentation included sharing real-world validation cases and various defense strategies. CyCraft emphasized that organizations should not solely rely on black-box solutions but rather develop robust, transparent mechanisms to prepare for new threats in the AI era.

Event Details

• - Date: November 18, 2025, from 14:25 to 14:55.
• - Venue: Bellesalle Takadanobaba Hall A (Track 2).
• - Speaker: Renata Chang, Japan Architect Consultant, CyCraft.
• - More Information: CODE BLUE 2025 Schedule

CODE BLUE 2025 Event Overview

The CODE BLUE conference runs from November 18 to November 19, 2025, taking place at Bellesalle Takadanobaba in Shinjuku. This annual event gathers security experts to discuss the latest in cybersecurity, particularly in the AI sector.

• - Official Site: CODE BLUE 2025

Related Innovations

• - CyCraft's presentation was just one of many initiatives aimed at strengthening AI security. The company recently had a research paper accepted at the prestigious NeurIPS 2025 conference, a testament to their commitment to advancing AI and machine learning technologies.
• - Additionally, CyCraft introduced XecGuard, a latest defensive measure against LLM vulnerabilities, and officially released XCockpit EASM, the leading solution in exposure attack surface management.

Conclusion

CyCraft stands at the forefront of AI security innovation, continually adapting to new challenges posed by technological advancements. Their participation in CODE BLUE 2025 not only highlights their proactive approach but also sets the stage for further developments in AI security. As we move forward, events like these play an essential role in uniting industry leaders and fostering collaboration to create safer AI environments.