セキュリティ

The Depths of AI Security: Preventing Data Leaks to IBM Bob—Building “Proactive Defense” with MCP and Custom Rules “Wait, did that prompt just contain confidential information…?” For an engineer immersed in development, this is a spine-chilling moment. AI assistants like ChatGPT and IBM Bob have become indispensable partners that expand a developer’s cognitive reach. However, the flip side of this convenience is the risk of unintended information leakage. In an era where powerful AI agents have deep access to local environments and internal documentation, security measures that rely solely on “human caution” have reached their limit. ...

Why Was the “Sanctuary” of Enterprise AI Breached? The Vulnerabilities Exposed by McKinsey’s “Lilli” and the Frontiers of Defense Integrating AI into the core of business has become a “prerequisite” for securing a competitive advantage. However, do we truly recognize the “structural gaps” lurking within these robust systems? Recently, news that “Lilli”—the proprietary AI platform deployed across McKinsey & Company, the world’s premier strategic consulting firm—was breached through red teaming (simulated attacks) sent a quiet but significant shockwave through the tech industry. ...