Samsung has restricted employee access to generative AI tools after concerns that sensitive company information could be exposed through them. The move covers company devices and the internal network, while the company works on safety guidelines and AI tools of its own.
Why Samsung drew a boundary
According to an internal memo reported by Bloomberg, Samsung is concerned that information entered into external AI services could be stored on outside servers, remain there indefinitely, or be shared with other users. The memo followed an alleged incident in early April in which Samsung developers entered internal source code into ChatGPT.
A Samsung spokesperson confirmed that the memo existed. The reported incident illustrates a practical challenge for businesses: employees may use generative AI to help with tasks, but entering internal material can put that material beyond the company’s direct control.
The concern is not limited to source code. Samsung’s restrictions also address personal and company-related information entered during personal use of ChatGPT and similar services. The policy therefore reaches beyond company equipment to what employees submit to AI tools themselves.
What the restrictions cover
The ban applies to generative AI services including ChatGPT, Bard and Bing Chat on Samsung-provided computers and mobile devices, as well as on the company’s internal network. Employees using such tools personally are also prohibited from entering personal or company-related information.
The memo warns that people who do not follow the rules could be fired. That consequence makes the restriction a workplace policy, not simply a technical block on particular websites or applications.
Samsung is reportedly developing a feature to block generative AI use temporarily. The pause is intended to remain in place until the company’s safety guidelines for the technology are ready. Those guidelines could allow use in ways that support employee “productivity and efficiency,” according to the report.
Internal tools as a possible path forward
A broad restriction can reduce the immediate risk of employees submitting sensitive material to outside services, but it also limits access to tools that may help with routine work. Samsung is working on its own AI tools for document translation and summarization, according to Bloomberg.
Those planned tools point to a more controlled approach: a company can decide which tasks AI should support and set rules for the information employees can use with it. The source does not specify when Samsung’s tools will be available or what data they will use, so the company’s plans remain a work in progress.
More broadly, large organizations may want to train their own AI systems and deploy them locally. Doing so could give them more knowledge of the training data and greater control over how a system is used. The source presents this as a potential direction for companies concerned about privacy, legal exposure and proprietary information.
The wider question for business AI
Samsung’s decision reflects a tension facing companies adopting generative AI: these tools may assist employees, but organizations also need confidence about what happens to the information entered into them. Controls on access, internal guidance and company-built tools are different ways to manage that tension.
The article also describes a separate concern: an AI system might generate material that is already proprietary to another company. That possibility can make control over training data relevant to businesses, alongside the handling of user inputs.
OpenAI is developing a business version of ChatGPT that, according to the source, would not store input data or use it to optimize AI. However, the source says OpenAI had not provided information about the data corpus used to train GPT-4. For companies, the question is therefore not only whether a tool is useful, but also whether its data practices and protections fit their needs.
Samsung’s temporary ban leaves room for future workplace use under guidelines that have yet to be developed. Its approach shows how a company can pause broad access while it considers the safeguards needed to use generative AI for business tasks.