Latest / The Digital Warfare Podcast / Digital Warfare Podcast Daily Brief #135 July 23, 2026
Transcript
- Digital Warfare: You're listening to the Digital Warfare Podcast. Today, we are covering a security problem that most organizations already have, but many still cannot see: shadow AI. Employees are adopting artificial intelligence faster than security teams can inventory it, approve it, or govern it. And the central issue is no longer whether organizations should allow AI. That decision has already been made by the workforce. The real question is whether security teams know which tools are being used, what information is being shared with them, and what access those tools have been granted. Here's what is happening. Employees create accounts for AI assistance, productivity tools, coding platforms, meeting applications, and specialized business services without involving IT. Some tools are used through a browser, others appear as features inside existing SaaS applications, and increasingly, AI agents, plugins, add-ons, and model context protocol servers connect directly to corporate systems. Each connection expands the organization's attack surface, but unlike a traditional software deployment, These tools may never appear in an approved application inventory. That creates a visibility problem. Security teams cannot protect data flowing through systems they do not know exist. An employee may paste customer information into a chatbot, upload an internal document for analysis, share financial data with an unapproved assistant, or authorize an AI integration to access email, cloud storage, collaboration platforms, or business records. The action may feel temporary, the exposure may not be. Once information enters an external AI service, The organization may lose control over where it is processed, retained, logged, or reused. And the risk extends beyond prompts. Modern AI adoption includes file uploads, browser extensions, SaaS integrations, autonomous agents, and MCP connections. These tools may receive broad permissions into Google Workspace, Microsoft 365, Cloud Platforms, Customer Systems, or internal databases. That means Shadow AI is not simply another form of shadow IT. It can become shadow access. The challenge for defenders is that surveys and policy acknowledgements are not enough. Employees may forget which tools they tested. They may not consider an AI feature inside an existing application to be a separate service, and new tools are appearing too quickly for security teams to track manually. The first requirement is continuous discovery. Organizations need to identify new AI accounts, applications, integrations, and user activity as they appear. That visibility should include approved tools, unapproved tools, usage by department. And connections between AI services and sensitive corporate systems. The second requirement is data flow monitoring. Security teams need to understand when employees are sharing personally identifiable information, credentials, secrets, financial records, source code, or sensitive files with AI platforms. Without that context, an organization may know an application exists but still have no idea what risk it creates. Third, review permissions. An AI tool connected to email or cloud storage should be assessed like any other privileged third-party application. What data can it read? What actions can it perform? Does it retain access after the user stops using it? And can that access be revoked centrally? Fourth, move governance into the employee's workflow. A policy stored on an internal portal will not stop risky behavior at the moment it occurs. Users need clear, immediate guidance when they attempt to use an unapproved tool, upload sensitive information, or grant excessive permissions. The objective should not be to block every experiment. It should be to redirect employees toward approved services before convenience becomes exposure. And finally, treat AI governance as an ongoing operational process, not a one-time project. The AI environment will change continuously. New applications will appear. Existing SaaS platforms will add AI functions, and agentic systems will request deeper access to enterprise data. The strategic lesson is simple Shadow AI is not coming, it is already here. And organizations that focus only on formal AI deployments will miss the tools, accounts, and integrations creating risk outside the approved perimeter. You cannot secure what you cannot see. And with AI, what you cannot see may already have access to your most sensitive information. This has been your Digital Warfare Daily Brief. Stay secure. We'll see you tomorrow.