Japanese researchers, telecoms work to combat deepfakes in hierarchical culture

Much as computer viruses started as an arcane problem and became something as common as household insects, deepfakes have grown from their debut in the shadows to a problem every enterprise (and, increasingly, every individual) must face.

In Japan, telecom firms have taken up the deepfake detection gauntlet to crack down on synthetic audio generated using AI tools. NHK reports that NTT East, a subsidiary of major telco Nippon Telegraph and Telephone (NTT), has teamed up with NABLAS, an AI research startup incubated at the University of Tokyo, to produce an app that analyzes voice frequency and background noise for telephone voice fake detection.

A translated article about the project in the ICT Economic News says the app is designed to detect fake voices under various conditions, and to monitor factors including voice format and call patterns. Fake audio triggers an alert.

Deepfake bosses exploit corporate habits

With its deeply entrenched culture of hierarchy and respect, Japan presents a unique challenge for deepfake detection, in that people are accustomed to trusting figures of authority. A recent piece in IT Business Today elaborates: “The scary part about deepfake attacks in Japan is how cleanly they slip into the country’s corporate habits. Attackers don’t just mimic a CEO’s voice or face for fun. They use polished audio or video to pose as top leadership, push urgent wire transfers, or squeeze out sensitive M&A details.” Moving fast, they exploit the tendency to do what your boss asks of you. “The hierarchy system makes employees hesitate before questioning someone above them.”

That said, Japan recognizes the problem, and is developing regulations for so-called AI technology that include clearer rules on synthetic content and stronger penalties for deliberate fakery.

“Alongside regulation, domestic innovation has become the real engine of progress. Japanese research teams are building detection models that understand local language quirks, speech rhythms, and facial structures.”

Work underway to address deepfake threat

NTT East is not the only telecom supporting ventures to address the deepfake crisis. A fund affiliated with telco KDDI, the second-largest mobile operator in Japan behind NTT, recently invested in audio voice cloning tool Resemble AI.

Likewise with academia. A research team working out of the Japan Science and Technology Agency (JST) developed a deepfake detection tool for synthetic video. According to an abstract, Synthetiq’s unique approach involved the research team employing an “adversarial competitive research” approach, “in which the researchers separated into the offensive side and defensive side and competed. Through such an approach, the research team enhanced the level of technologies for safety enhancement in speaker authentication systems, protection of privacy, and automatic protection against voice impersonation attacks.”

Meanwhile, in April 2025, industry leader Reality Defender shared the top prize for the 2025 US-Japan Global Innovation Challenge, sponsored by the Defense Innovation Unit (DIU) and Japan’s Ministry of Defense Acquisition, Technology, and Logistics Agency (ATLA) and focused on combating information warfare.

Article Topics

AI fraud  |  deepfake detection  |  deepfakes  |  fraud prevention  |  generative AI  |  Japan  |  regulation  |  synthetic data  |  voice biometrics