SB OAI Japan brings Patching as a Service to 3,000 firms

SB OAI Japan was set up roughly a year ago with a stated goal of building "enterprise AGI" — a world where AI moves across all of a company's systems and data. SoftBank serves as its first test bed; under chairman Masayoshi Son's top-down direction, the venture has run internal use-case contests to surface ideas from staff. The bet is that OpenAI supplies the model and training layer while SoftBank handles introduction support — including system integration and management-side change — a division of labor that reflects the view that US best practices do not transfer directly to Japanese firms.

The security push took shape after Anthropic reportedly used "Claude Mythos" around spring 2026 to surface many vulnerabilities in web browser apps and OSes. CCSO Kengo Kanazawa says SB OAI Japan ran its own diagnosis and found vulnerabilities it had missed until then, and warns that current defenses would not withstand attackers using AI. The service relies on autonomous reasoning models rather than the pattern-matching approach long standard for vulnerability scans.

The longer-term aim is legacy modernization: most Japanese companies run custom systems built with SIer help, and Kanazawa argues that fixing individual bugs invites new ones as long as hand-built systems sit on old infrastructure. SB OAI Japan appears to see modernizing that infrastructure as a prerequisite for AI agents to cooperate across a business, which is the premise of the enterprise-AGI vision the venture was founded on.