#Japan #Tokyo #AI Agents #Safety Platform #DataGusto

DataGusto Launches its Innovative AI Safety Platform

On December 11, 2025, DataGusto, a Tokyo-based startup dedicated to AI safety, introduced its groundbreaking platform aimed at preventing unexpected behaviors in AI agents, often referred to as "malfunctions". The platform is designed to evaluate the reliability of AI systems and ensure their safe operation. As businesses increasingly adopt AI technology for workflow automation, the risks associated with AI agents have become a growing concern.

The Challenges of AI Agent Safety

In the past, businesses have relied on static guardrails—pre-defined rules that dictated the acceptable behavior of AI agents. However, as the number of deployed AI agents has surged, the need for constant updates to these guardrails has become apparent. The traditional method struggles to keep up when new models are implemented or when agents evolve, leading to the potential for errors that were previously deemed manageable. A prominent example is an incident reported overseas, where an AI coding platform inadvertently wiped a live database due to a misjudgment that traditional monitoring systems failed to prevent.

As companies undergo the transition from mere AI exploration—searching and generating content—to more complex AI-driven automation, the pressing need for a reliable safety mechanism grows. DataGusto's solution focuses on real-time detection and prevention of risks associated with AI agents, utilizing advanced technology that transcends the limitations of static guardrails.

Real-Time Risk Management

Recognizing a gap in traditional risk management approaches, DataGusto devised a novel methodology that enables dynamic creation of guardrails in real time. This means that the platform can actively adjust to new circumstances as they arise, establishing boundaries for behaviors judged safe or unsafe at any given moment. It’s a proactive approach that addresses both existing and unforeseen risks, thus reinforcing the reliability of AI agents.

Independent Evaluation of Safety

In a confirmation test using the "PropensityBench" benchmark from Scale AI, DataGusto demonstrated a 100% prevention rate for incidents that static measures historically failed to address. This groundbreaking result positions DataGusto as a pioneering force in not just safeguarding against external threats, but also managing internal risks associated with AI behavior—an area often overlooked in conventional safety assessments.

The Importance of Safety Measures

As the use of AI agents becomes more prevalent, especially within complex ecosystem environments, the necessity for robust safety measures increases. It’s not enough to safeguard against cyber threats; businesses must also take steps to ensure that AI outputs are consistent with human intent, thus fostering trust in AI systems. The Ministry of Internal Affairs and Communications indicates that nearly 50% of companies are establishing policies related to generative AI utilization as of 2024, highlighting the urgency of addressing safety risks in AI development.

Future Prospects and Multi-Agent Systems

Looking ahead, the proliferation of multi-agent environments where various AI systems interact will necessitate a foundational layer of trust—a role that DataGusto aims to fulfill. Our goal is to not only ensure safety in individual agent operations but also to create a reliable system of interconnected agents that can autonomously manage societal functions.

Join Us in Enhancing AI Safety

DataGusto is currently seeking partner companies interested in improving safety in AI agent operations. If your organization is looking to incorporate advanced risk management protocols into AI initiatives, please reach out to us. Together, we can pave the way towards a future where AI agents can operate safely and effectively, transforming industries and society at large.

About DataGusto

Founded with the vision of enhancing the reliability and safety of autonomous AI agents, DataGusto is spearheaded by CEO Mao Pa, an expert in AI governance consulting. The company aims to provide a safety platform that prevents unintended incidents arising from AI behavior. With a strong focus on integrating advanced safety measures, DataGusto is at the forefront of innovation in the AI field, ensuring that the benefits of AI can be realized while minimizing risks.