Tokyo-based MYUUU has introduced KUNADO, a proprietary AI model and agent built specifically for cybersecurity. The system examines web applications, APIs, cloud configurations, and core business systems from the perspective of an attacker. It autonomously surfaces exploitable weaknesses and ranks them by priority, giving organizations a clearer view of their most pressing security gaps.
A Shifting Threat Landscape
Cyber attackers are increasingly using artificial intelligence to automate vulnerability discovery, lowering the cost and effort required to launch campaigns. MYUUU points to an IBM survey indicating that the share of malicious data breaches involving AI-assisted attacks rose by 56% compared with the prior year. As a result, smaller companies and long-running internal systems that once considered themselves unlikely targets now face realistic exposure.
How the AI Agent Works
Unlike conventional tools that compare systems against static checklists, KUNADO forms hypotheses about where an environment could be compromised. It then executes tests, validates results, and refines its approach in a continuous loop that mirrors the behavior of a skilled human hacker. This method helps uncover risks that emerge from combinations of specifications, business logic, and configuration choices rather than isolated known flaws.
Security and Deployment Controls
MYUUU runs KUNADO's proprietary model inside inference infrastructure located in Japan, which keeps source code and diagnostic data from being sent to overseas AI services. The agent operates within agreed boundaries for targets, permissions, and actions, with a read-first approach and pre-established stop conditions to limit production impact. The company notes that this design reduces operational risk, although it does not guarantee a completely non-destructive assessment.
Findings From Real World Assessments
KUNADO and MYUUU's diagnostic support have repeatedly uncovered issues that bring organizations close to an incident. In one case, a legacy core system used an unsupported software version and transmitted customer information over an unencrypted HTTP connection. In other cases, a reservation app created with generative AI exposed customer data from other stores through an API due to an access control gap, and a website contact form plugin allowed external parties to view submitted business information.
Reporting and Remediation Guidance
Every identified issue is compiled into a monthly report that explains the supporting evidence, the potential impact of exploitation, and the priority for remediation. The report also includes specific remediation steps so teams know where to begin. MYUUU positions this output as a way to turn security findings into a clear action plan rather than a long list of alerts.
Leadership and Use Cases
Representative Director Togo Muroya said that information leakage has become a direct management risk, with disclosure events leading to lower stock prices and shareholder litigation in some cases. He argues that if attackers use AI to find weaknesses, defenders should adopt the same perspective while controlling impact and acting proactively. MYUUU sees KUNADO as suitable for pre-launch product checks, periodic reviews of core systems and CRM platforms, and validation of internally developed applications.
MYUUU was founded in February 2025 and is headquartered in Shibuya, Tokyo, with a focus on generative AI and web services alongside security diagnostics. The company plans to expand the range of supported systems and integration options, covering both pre-release verification and continuous monitoring. Organizations concerned about public-facing products or aging infrastructure can contact MYUUU through the KUNADO service website at kunado.ai.