KEY POINTSMMRI survey finds 60% of Japanese corporate security decision-makers feel threatened by more advanced AI-driven cyberattacksOnly 13% say current security systems can fully handle AI-enabled attacks, while 81% report at least partial gapsOn-premises security AI draws 53% interest as 63% express concern over cloud-based specialized AI
A survey by MM Research Institute shows 60% of firms fear AI-driven cyberattacks. Note: Japanese text in the original image from press release has been translated into English by our editorial team.
MM Research Institute
Japanese research firm MM Research Institute, or MMRI, said on August 27 that a survey of 1,000 corporate security product selectors and security operations managers found growing concern over more sophisticated cyberattacks using artificial intelligence, alongside stronger interest in domestically developed, on-premises AI tools.
The web-based survey was conducted from July 17 to July 20, 2026. It found that 60% of respondents said their companies felt a sense of crisis over the increasing sophistication of AI-enabled cyberattacks. When asked how well their current security structures could respond to such attacks, only 13% said they could respond sufficiently, while a combined 81% said they might be unable to respond in part or could hardly respond at all.
Among perceived threats from AI-enabled attacks, the most frequently cited was optimization of ransomware target selection and attack tactics at 42%. High-precision phishing and spear-phishing at scale and automated discovery and exploitation of unknown vulnerabilities, or zero-day flaws, followed at 37% each.
When asked where AI would be most useful in their own security systems, 42% chose defense and 21% chose detection, making defense roughly twice as popular as detection. The most sought-after specific function was checking and strengthening security settings, or hardening, at 46%, followed by attack interception and blocking at 43%.
The survey also pointed to concerns over cloud-based, security-focused AI tools. MMRI said 63% of respondents had concerns about using cloud-based specialized AI for vulnerability diagnosis. The top reason, cited by 44%, was the possibility that source code or corporate data could leak externally through the AI, while 42% pointed to the act of sending such code or data to outside AI tools. Concern about data being processed or stored on servers outside Japan stood at 36%.
Dependence on foreign AI vendors was another issue, with 67% expressing concern. At the same time, 53% said they were interested in on-premises specialized security AI, and among that group, 89% said they would be interested if such systems were developed by Japanese companies.
MMRI said larger companies were more likely to feel a sense of crisis about more advanced AI-driven cyberattacks and more likely to say they could respond to them. Respondents in financial services, including banking, securities and insurance, as well as telecommunications, also showed relatively high levels of concern and confidence in their response capabilities.
The release referred to cloud-based specialized security AI such as “Claude Mythos” as an example of a category that raises concerns for some users. It also said hybrid deployment models could become an option as companies seek to improve defenses while retaining tighter control over sensitive code and data.
Cybersecurity has become a broader operational issue in Japan as attacks on large companies have caused system disruptions affecting suppliers and customers as well as the companies directly targeted. The survey results suggest companies may be showing increased interest in security products that combine AI-based defense functions with domestic data handling and on-premises deployment.