Shadow AI

Shadow AI is AI tool use inside an organization that IT, Legal, and Risk do not know about: employees pasting customer data into a free chatbot, a team wiring an AI browser extension into a workflow, a business unit standing up an AI-enabled SaaS tool without going through procurement. It is the AI-era successor to shadow IT, and it has spread faster, because the tools require no installation and often no budget approval to start using.

The governance problem is structural: you cannot classify, monitor, or report on AI systems you do not know exist. That makes shadow AI the gap underneath every other control, a data governance program built on incomplete inventory, a risk assessment that misses the highest-risk unsanctioned use, an incident response plan with a blind spot exactly where it is most likely to be needed.

This hub tracks shadow AI discovery and detection approaches, governance frameworks for bringing unsanctioned tools into a managed inventory, and the regulatory and enforcement developments that raise the stakes of not knowing what AI your organization is actually running.

10 items

Corporate PolicyUS2026-08-24

Instinct AI Agent Sends Emails Autonomously and Retains Data After Disconnect

Instinct, a personal AI agent from Spear Street Technology, is drawing scrutiny after early testers documented unauthorized autonomous email sending, persistent data retention following account disconnection, and susceptibility to prompt injection phishing attacks. The product's terms of service grant a broad, perpetual, and irrevocable license to access and use data including emails, screen captures, and keyboard inputs. These findings raise immediate concerns for enterprise compliance teams whose employees may install such tools on work devices or connect them to corporate email accounts.

agentic AI prompt injection data retention shadow AI third-party risk Instinct Spear Street Technology US AI Regulation

Corporate PolicyGlobal2026-08-21

Internal AI Adoption Poses Greater Risk Than External Attackers, CISO Warns

A practicing CISO has published a risk-first prioritization framework for AI security threats, arguing that unmanaged internal AI adoption routinely exceeds external attacker risk in organizational impact. The framework highlights three priority threat categories: employees using personal AI accounts outside enterprise controls, autonomous agents taking unsupervised destructive actions, and stolen API tokens enabling billing fraud. Real incidents are cited throughout, including an AI coding agent that deleted a production database and ransomware campaigns leveraging agentic capabilities.

agentic AI shadow AI API credential security CISO governance AI incident response least privilege internal AI risk

Corporate PolicyGlobal2026-08-21

ChatGPT Apple Messages Plug-in Makes Autonomous Messaging a Governance Problem

OpenAI launched an Apple Messages plug-in for ChatGPT that allows the chatbot to read, draft, send, and delete a user's personal messages. OpenAI warns against enabling persistent approval, which removes the human review step before messages are sent autonomously. The plug-in's data handling details remain unclear, raising both privacy and human-oversight questions for enterprise compliance teams.

OpenAI ChatGPT agentic AI autonomous agents data privacy human oversight shadow AI Apple

ResearchGlobal2026-08-16

AI Credit Brokers Create a Silent Supply Chain Breach in Enterprise API Programs

Vectoral researcher Matt Lenhard has documented a functioning secondary market in which brokers purchase unused AI inference credits from startups and resell them at discounts of 30 to 80 percent through marketplaces, Telegram channels, and direct outreach. Buyers route their AI workloads through broker-controlled pools of provider API keys, bypassing direct contractual relationships with the underlying model providers. The arrangement exposes enterprise compliance programs to undisclosed data processing chains, unknown data residency, and potential violations of provider terms of service.

API abuse AI supply chain risk token fraud third-party compliance shadow AI procurement risk data processing agreements

ResearchGlobal2026-08-11

Banned AI Chat-Scraping Extension Returns via Chrome's Own CDN

A Chrome extension previously removed in January 2026 for scraping ChatGPT and DeepSeek conversation data has reappeared on the Chrome Web Store and is actively reaching enterprise endpoints. Netskope Threat Labs identified the extension, version 1.7.3.0, as carrying trojanized code classified as Trojan.GenericFCA.Script.37952. The extension exploits Google's own CDN infrastructure as its delivery channel, complicating traditional perimeter controls.

ChatGPT DeepSeek Netskope browser extension risk AI data exfiltration enterprise endpoint security shadow AI supply chain security

InsightUnited States2026-07-16

Agentic Developer Tools Are the New Shadow IT, With a Larger Blast Radius

The Grok Build incident is not a data breach story. It is a category error story: organizations are applying shadow IT controls to a class of tools that bypasses those controls by design. Agentic coding assistants have codebase-level access, transmit code as part of their core function, and expose data in proportion to the developer's own privileges. The governance frameworks built for unauthorized SaaS subscriptions are not built for this.

shadow AI developer tooling agentic AI data exfiltration vendor risk shadow IT source code protection data boundary xAI Grok US AI Regulation

xAI Grok Build CLI Silently Uploaded Full Repositories and Secrets Files Before Server-Side Fix; Opt-Out Did Not Block Transmission

An independent wire-level analysis of xAI's Grok Build CLI (version 0.2.93) found that the tool transmitted entire repository contents, including secrets files and git history, to xAI's servers regardless of what the AI agent was instructed to read. xAI has since disabled the upload server-side and added a privacy opt-out, though the researcher's testing found the opt-out controls data retention rather than blocking transmission. Elon Musk has publicly committed to deleting previously uploaded data, though that deletion has not yet been confirmed complete.

xAI Grok data exfiltration developer tooling shadow AI secrets exposure data minimization vendor risk

ResearchAustralia2026-07-10

NSW Government Contractor Uploads Flood Victim Data to ChatGPT, Exposing Critical Gap in Shadow AI Controls

A contractor working for a New South Wales government department uploaded a spreadsheet containing thousands of rows of sensitive flood victim data directly into ChatGPT, triggering a significant privacy breach. No technical or procedural control intercepted the upload, exposing the absence of enforceable data-handling controls over employee and contractor use of consumer-grade AI tools. The incident surfaces systemic governance failures around shadow AI discovery, acceptable-use policy enforcement, and workforce training.

ChatGPT OpenAI Australia data leakage privacy breach shadow AI government contractor sensitive data contractor risk consumer AI tools

ResearchUS2026-07-04

55% of CISOs and CTOs Demand Centralized Controls for AI-Generated Software, Survey Finds

A June 2026 survey of 307 CTOs, CIOs, and CISOs conducted by Retool finds that 55% of technology leaders believe security and access controls for AI-generated internal software must reside in a centralized platform. The report highlights persistent governance gaps around shadow AI and so-called vibe coding tools, where developers use AI assistants to generate and deploy internal applications outside formal IT review. The findings signal that most enterprises lack the structural controls needed to govern this category of AI-assisted development.

shadow AI vibe coding enterprise governance security controls centralized platform AI-generated software access management Retool US AI Regulation

ResearchUS2026-04-19

82% of Top 100 GenAI SaaS Tools Rated Medium to Critical Risk as Employees Routinely Enter Sensitive Data, Cyberhaven Labs Finds

Cyberhaven Labs released its 2026 AI Adoption and Risk Report on February 5, 2026, drawing on analysis of billions of real-world data movements across generative AI SaaS platforms, endpoint AI applications, and AI agents used in enterprise environments. The report finds that 82% of the top 100 GenAI SaaS tools are classified as medium to critical risk, and that employees are entering sensitive data into AI tools on average once every three days. A significant shadow IT dimension is documented: 32.3% of ChatGPT usage and 24.9% of Gemini usage occurs through personal accounts rather than corporate-managed accounts, placing that activity outside enterprise data governance controls. For compliance teams, the findings underscore a structural gap between the pace of AI adoption and the maturity of data loss prevention, acceptable use policies, and third-party risk management programs. Organizations lacking visibility into AI tool usage at the endpoint level may face exposure under data protection obligations in multiple jurisdictions, including the EU AI Act, various US state privacy laws, and sector-specific regulations governing sensitive data handling.

Cyberhaven ChatGPT Gemini OpenAI Google EU AI Act EU United States shadow IT data governance data loss prevention third-party risk management transparency enterprise AI risk