Shadow AI
Also called: shadow IT for AI, unsanctioned AI tools
Shadow AI is the use of AI tools, models, or services inside an organization without IT, security, or governance team approval. It mirrors the broader concept of shadow IT but specifically covers generative AI assistants, agents, and ML services.
Most organizations underestimate their AI footprint by 3-5x. Employees adopt new AI tools faster than procurement, security review, and policy authoring can keep up — sometimes within hours of a tool launching. Common shadow AI categories include:
- Browser-based generative AI (ChatGPT, Claude, Gemini, Perplexity) used through personal accounts
- Coding assistants (Copilot, Cursor, Codeium) with auto-suggest enabled in private repos
- Productivity AI (Notion AI, Otter, Fireflies) embedded in adjacent tools
- Browser extensions that quietly route page content to AI APIs
- AI-powered features inside otherwise sanctioned SaaS that activate when "AI" is toggled on
Each shadow AI use case carries a distinct risk profile. A coding assistant trained on customer data is a regulatory disclosure event. A meeting transcription bot may capture privileged communications. A model that retains prompts indefinitely creates an indefinite breach surface.
Detection requires more than asking employees what they use. Self-reports miss what people forget, what they think is harmless, and what they don't want to disclose. The reliable signal is at the network layer: TLS handshakes, DNS resolutions, and request patterns reveal which AI services are actually being contacted, when, and from where.
Why it matters
Shadow AI is the gap between your AI policy and reality. Compliance frameworks (NIST AI RMF, ISO 42001, EU AI Act) all assume you know which AI systems are in use. Without that inventory, you can't classify risk, write meaningful policies, or attest to controls. Discovery has to come first — every other governance artifact builds on it.
Related terms
AI Governance
AI governance is the set of policies, processes, roles, and controls an organization uses to develop, deploy, and operate AI systems responsibly and in compliance with applicable laws, standards, and stakeholder expectations.
AI Bill of Materials (AI BOM)
An AI Bill of Materials (AI BOM or AIBOM) is a structured inventory of every component used to build, train, and deploy an AI system: training data sources, foundation models, fine-tuning datasets, prompts, libraries, dependencies, and downstream integrations.
Trust Gap
The Trust Gap is the difference between an organization's self-reported AI security posture and the posture verifiable from independent evidence.