SoftBank Partners with OpenAI to Launch AI Cybersecurity Services in Japan

Masayoshi Son, Chairman of SoftBank Group [Photo=Yonhap News]

SoftBank Group is set to launch a corporate cybersecurity service in Japan utilizing OpenAI's artificial intelligence (AI) models. This announcement comes shortly after the U.S. government restricted access to a competing AI model specialized in cybersecurity from Anthropic, indicating a strategic move to address the demand for AI defenses among Japanese businesses and public institutions.

According to Nikkei Asia on June 16, SoftBank will offer a security service aimed at Japanese companies that leverages OpenAI's high-performance models to identify vulnerabilities in corporate systems. The company plans to expand the service to include threat response and system improvements in the future.

Previously, Anthropic halted user access to its robust cybersecurity AI model, Claude Mythos, following U.S. government actions. This raised concerns among Japanese companies and government agencies that intended to use the model for cybersecurity.

Masayoshi Son, Chairman and CEO of SoftBank Group, announced at an event in Tokyo that the service will be rolled out to approximately 3,000 critical infrastructure operators in Japan. SoftBank aims for a full launch by the end of this year.

Son warned that AI-driven cyberattacks could represent a historical turning point for Japan. He referenced the 19th-century incident involving U.S. Commodore Perry's fleet, which pressured Japan to open its ports, stating, "AI cyberattacks could be the biggest crisis Japan has faced since the Black Ships incident."

He predicted that malicious organizations and attackers would increasingly use AI to launch large-scale assaults. He noted that critical infrastructure operators in sectors such as utilities, finance, telecommunications, and transportation face heightened risks as AI reduces the cost and increases the scale of attacks.

SoftBank also plans to expand its cybersecurity support staff from about 50 to 1,000. The company secured access to OpenAI's cybersecurity-focused model, GPT-5.5-Cyber, at the end of May and reported identifying over 10,000 vulnerabilities after assessing 700 internal systems.

However, SoftBank has not disclosed which model will be applied to the new service or whether GPT-5.5-Cyber will be used. The trend of restricted access to AI models in the U.S. has created uncertainty regarding the utilization of security technologies in Japan, meaning that OpenAI-based services may also be affected by future policy changes.

Sam Altman, CEO of OpenAI, highlighted in a video message that cybersecurity is one of the most critical applications of advanced AI. He stated, "OpenAI's goal is not just to help organizations find vulnerabilities but also to assist them in strengthening their systems and solving complex security challenges."

* This article has been translated by AI.