Fujitsu sees opportunities in AI to tackle social issues. These range from digitizing healthcare policy to simulating cyberattacks in a virtual network environment. We spoke with Dr. Kurt Dusterhoff, Senior Manager, Technology Strategy Unit.
How do you ensure that AI doesn’t remain just a technological toy, but also delivers real societal value? That is the question that keeps Dusterhoff of Fujitsu busy every day. The Japanese technology company is working on applications for marine ecosystems, healthcare, and cybersecurity, each with a common thread: data-driven decision-making with a human overseeing the process.
One of those domains is the marine world. Last year, Fujitsu developed an Ocean Digital Twin to map seaweed and marine algae. This includes their ability to absorb CO2 as part of a “Blue Carbon economy,” which focuses on the financial and ecological value of carbon storage by marine and coastal ecosystems. The project is currently running primarily along the Japanese coast, and Dusterhoff is in talks with potential clients in Europe and other regions. Collaboration is taking place with local governments, port authorities, and marine companies.
Tailored healthcare policy via digital twin
Fujitsu is currently achieving more concrete and visible results in healthcare. Fujitsu applied its digital twin approach to policy simulations for kidney damage in diabetes patients. The problem with existing healthcare policy is that everyone follows the same care pathway, regardless of age or demographic background. This leads to inefficiency and sometimes results in people who need a different approach being reached too late.
Dusterhoff explains how Fujitsu tackles this. AI converts policy documents into uniform, digital flowcharts, enabling the simultaneous simulation of dozens of policy variants. The system then compares the variants to determine which best aligns with the local population composition. “People deliver healthcare. Policy flows do not provide healthcare,” states Dusterhoff. The workflow merely determines when, who, and what intervention is provided. This could be a letter, a text message, a phone call, or an invitation to see the family doctor.
In trials in Japan, this approach yielded twice as many positive outcomes at half the cost as the existing method. This is not because the AI makes fewer mistakes than humans, but because targeted early intervention prevents later costs.
Human oversight remains essential in this system. When 80 percent of the notifications sent generate a response, the care team focuses on the 20 percent that do not respond. The AI filters out the noise and makes the workload manageable, but healthcare professionals continue to make the decisions.
AI agents as a weapon and shield in cybersecurity
The second major application is cybersecurity. And that world is changing rapidly. Botnets continuously scan for vulnerabilities. As soon as a breach is found, an attacker is ready with off-the-shelf AI tools to exploit it. The barriers—cost and technical knowledge—have virtually disappeared.
Fujitsu’s solution is a simulation environment that combines a corporate network with known vulnerabilities. As soon as a new vulnerability list becomes available, Fujitsu can launch a subnetwork within 15 to 20 minutes that specifically targets that vulnerability. Running twenty to thirty such simulations simultaneously is no problem. The system executes the attack, tests mitigations, and ultimately returns a prioritized list of what needs to be addressed in the real network.
After that, it’s up to the human. Because the best mitigation involves restarting all servers, someone with operational knowledge of the company must decide when and how to organize it.
In addition to attack simulations, Fujitsu is also working on the other side: securing AI systems themselves. The company has developed an LLM vulnerability scanner with over 9,000 known vulnerabilities, ranging from jailbreak techniques to prompt injection. This scanner operates on top of the standard guardrails of hyperscalers like OpenAI, Google, and Meta, and adds company-specific logic. This allows organizations to configure which departments are permitted to ask which questions of which AI system and to translate their own policy rules into technical guardrails.
Fujitsu’s cybersecurity teams operate from Japan, the United Kingdom, and Fujitsu Research of Europe in Israel, which collaborates with Ben-Gurion University. Dusterhoff describes his role within this framework as “the human glue”: a connector between teams that may not always be on the same page, but who are working on the same problems.
Tip: Fujitsu, Intel and Qualcomm collaborate under auspices of Linux Foundation