Employees Feed Internal Data Into Unapproved AI Tools as Security Gaps Widen
Published on 07/13/2026 at 12:54 | Redaktion boerse-global.de
A new study suggests that a significant portion of the German workforce is using artificial intelligence tools without their employer’s knowledge or consent, raising concerns about data leaks and compliance. Conducted by Bilendi and released in June 2026, the survey found that 42.7 percent of employees who use AI systems have never obtained formal clearance from their companies.
The unauthorised activity — often called "shadow AI" — spans a broad range of sensitive information. More than two in five users (42.7 percent) admitted to pasting drafts of internal emails into public AI platforms. 15.7 percent shared strategic company information, while 12.9 percent processed confidential customer data. Over 60 percent of all respondents said they want clearer rules and targeted training on proper AI use.
The gap between employee behaviour and corporate policy is widening faster than many organisations can respond. According to the Digitalization Survey 2026 published by the German Chamber of Industry and Commerce (DIHK), only 35 percent of companies currently use AI. Another 34 percent have plans to adopt it within the next three years. The biggest obstacle cited was a lack of acceptance and a need for further training, named by 31 percent of firms — topping even the well-known shortage of IT specialists.
Roughly two-thirds of companies surveyed said they feel dependent on AI vendors, a situation that compounds the risk when employees go off-book. The discrepancy between rapid tool adoption and inadequate oversight is growing, the DIHK report notes.
Regulators are beginning to tighten the rules. Germany’s financial supervisor BaFin specified new responsibilities in December 2025 under the three-lines-of-defence model for AI. The authority warned that structural gaps emerge when operational implementation and internal control are not kept separate. It recommended a three-phase approval process consisting of preliminary inquiry, validation and monitoring.
On 11 July 2026, the United States Supreme Court issued a ruling that sharpens liability for freight brokers, increasing demands for documentation and verification of subcontractors. Legal experts reacted by calling for robust review procedures and regular training to meet the new standards.
Meanwhile, authorities in other regions are pushing forward with digital compliance tools. On the same day, the Securities and Exchange Commission of Pakistan introduced a digital know-your-customer (KYC) process based on IBAN verification. The system aims to cut onboarding times and reduce paperwork.
Software providers are responding to the wave of regulation. Pega has started embedding compliance rules for anti-money laundering (AML), MiFID II and the Common Reporting Standard (CRS) into low-code platforms, with interfaces to databases such as World-Check for continuous monitoring. Wolters Kluwer launched an AI assistant for its ADDISON software that delivers automated answers on bookkeeping and tax matters.
The technology itself is evolving beyond basic chatbots. In 2026, AI agents capable of independently pursuing goals are entering the workplace. One example is "Xero Ultra," released on 11 July 2026, which uses an AI agent for accounting and cash-flow forecasts. Separately, Microsoft is replacing some OpenAI models in Excel and Outlook with its own MAI series, aiming for tighter control and efficiency.
Global market analysts forecast artificial-intelligence investments exceeding $3 trillion by 2027. In corporate boardrooms, a majority of executives expect restructuring and job losses, yet many also see AI as a measurable driver of productivity.
Disclaimer regarding our articles: No investment advice, no buy or sell recommendation. Information on prices, companies, and markets is provided without guarantee; changes are possible at any time. Stock market transactions can lead to substantial losses. Our articles are created and reviewed in whole or in part automatically with the support of AI.
