Trust in AI: the starting point for transformation
— Co-creating a sustainable future —

1. Explainability: Charting a path to shared understanding

When a company asks AI how it reached a certain conclusion about an important decision, the answer must be unambiguous. Otherwise, they cannot accept the AI’s decision or take responsibility for them. Therefore, being explainable is a prerequisite for transforming AI from something people simply trust into something they can confidently rely on and entrust with real decisions.

Fujitsu is pursuing technologies that make AI's decision-making processes transparent. For example, our Knowledge Graph, which systematically relates vast amounts of knowledge, visualizes AI's thought process (i.e. logical reasoning and basis for output) as a knowledge map. Furthermore, RAG technology*1, built around this Knowledge Graph, dramatically enhances the accuracy of the data that serves as AI’s knowledge repository, allowing it to find accurate evidence from vast internal regulations and past expertise.
Causal AI*2 clarifies the "why" behind events. Is the true cause of declining sales performance market fluctuations or a misstep in promotion? It objectively identifies causes and effects, not just correlations, ensuring reproducibility in business and allowing businesses to confidently take the next strategic action.

2. Security & Ethics: Enabling collaboration with people and a fairer society

Against the backdrop of AI proliferation, demand is strong for preparedness against vulnerabilities, cyber threats, and ethical risks. Being secure and ethical is a crucial prerequisite for companies to earn trust from society and achieve sustainable growth.

In 2019, Fujitsu formulated the Fujitsu Group AI Commitment. By embracing an “Ethics-by-Design”, we integrate ethics from the design stage. This approach includes measures such as establishing a committee composed of external experts. We are also a founding member of AI4People, Europe's largest AI ethics organization, and we continue to participate in international forums such as the OECD and AISI, actively engaging in rulemaking itself. In Japan, we also participate in discussions on revising the Ministry of Internal Affairs and Communications and the Ministry of Economy, Trade and Industry’s “AI Guidelines for usiness,” and we have also taken part as an expert committee member in discussions during the formulation of Japan’s AI legislation.

We are also engaged in technological development to address cyber threats unique to the AI era. Multi-AI agent framework*3 enables AI to discover vulnerabilities through repeated attacks and defenses, and to derive countermeasures. This enables proactive security measures, even without specialized security expertise.

3. Sovereignty: Ensuring autonomy of data and AI

The importance of data and AI sovereignty has become greater than ever. In modern management, it is essential that data is appropriately managed and that AI systems themselves function autonomously without undue external influence.

Fujitsu defines sovereignty from three perspectives: Security, Flexibility, and Domain specificity (along industry and business lines). The only way for an organization to place AI at the core of their business and unleash its true value is to act on these areas and guarantee the right to control their own data and AI.

Takane*4, an enterprise-grade large language model (LLM) co-developed with Canadian AI startup Cohere, is characterized by its use in secure private environments. The multi-AI agent framework*5 tailors AI agents for specific domains and autonomously improves workflow quality in secure environments. It also optimizes collaboration among agents while they grow in a specific domain, while managing interactions between agents securely and with a high degree of relevancy.

Secure Agent Gateway*6 enables AI agents from different companies to collaborate without leaking confidential information. Leveraging this and related technologies, we aim to forge “AI spaces,” collaborative environments where multiple agents can coordinate, learn, and connect through data.

4. Sustainability: Sustainable operations and a flexible foundation

As AI evolves and adoption rates rise, the energy consumption and costs required for computation are increasing exponentially. This is a management challenge that no company can afford to ignore, from both sustainability and financial discipline standpoints. For AI to become trustworthy over time, it must remain sustainable to operate as adoption scales.

FUJITSU-MONAKA*7, a next-generation CPU scheduled for commercialization in 2027, significantly improves power efficiency for AI processing. We are also working to achieve both energy savings and high efficiency through strategic collaborations.*8

We are also developing an AI compression technology for LLMs*9 that will enable large AI models to run efficiently with fewer resources, dramatically reducing the enormous computational resources that would normally be required. Furthermore, by combining high-performance computing technologies cultivated through years of supercomputer development with quantum computing technologies that will deliver the breakthroughs of tomorrow, we are taking a multi-faceted approach to AI sustainability.

5. Connectedness: Unlocking new value through networks and collaboration

Harnessing the true power of AI isn’t possible if utilization is confined to specific departments or business processes. True transformation emerges from the collaboration of data and AI across organizational and industrial boundaries. However, it is also true that connecting different systems and data securely and without constraints presents difficulties.

For that reason, Connectedness is not merely about technical interfaces. It is an innovation catalyst that breaks down organizational silos and enables cross-industry collaboration. For example, AI-RAN aims to create new services through real-time data analysis. Key AI technologies such as Fujitsu Kozuchi*10 and Takane are designed with high connectivity at their core. Connecting knowledge across organizational and corporate boundaries and expanding the framework of new business value—that is the advantage of the transformation brought about by connectivity.

The combination of these five elements creates trust in AI

Now that we have introduced the five elements individually, it’s important to note that the most important thing is that these are not disparate functions; unwavering trust is only born when they function as a single, integrated system. Accuracy is meaningless without safety, and safety still leaves risks if sovereignty is absent. These structural requirements form the foundation that supports the positive cycle of industrial transformation and growth presented in the next section.

These are just a few examples of the transformation and growth that Trust in AI can bring. In an era of increasing uncertainty, clinging to past successes or settling for partial AI adoption can lead to missed growth opportunities and serious strategic risks for companies. By optimizing connections and integrating the five elements suited to individual challenges, AI goes beyond simply being usable to being reliable, making it a true engine of transformation.