Samsung said on June 9 that it will roll out external generative AI tools — Google's Gemini, OpenAI's ChatGPT, and Anthropic's Claude — across all of its affiliates this month, the first time the conglomerate has adopted outside AI models companywide. The reversal is the story: this is the same company that banned employees from using public generative AI in 2023 after engineers leaked confidential data into ChatGPT. Samsung is now re-admitting those tools — but only behind a security control layer it built specifically to prevent a repeat.
The push, which Samsung calls the starting point of an "AI transformation," or AX, aims to embed AI across the entire value chain, from research and development and manufacturing to marketing and support. Samsung framed AI not merely as a productivity tool but as a method for fundamental management change and a means of finding new growth.
The initiative traces directly to Samsung Electronics Chairman Lee Jae-yong, who used his New Year's message in January to say the company must fundamentally change the way it works and its organizational DNA, and incorporate AI across the whole business. A Samsung official framed the effort as the next in a line of pivots — after digital and mobile transformations — and the starting point for becoming an "AI-native company."
Why Samsung Banned These Same Tools in 2023
The security gating is not boilerplate caution; it is a direct response to Samsung's own history. In March and April 2023, within weeks of allowing ChatGPT on its semiconductor operations, Samsung engineers pasted proprietary source code, an equipment-defect sequence, and a confidential meeting transcript into the chatbot. Because consumer AI services can retain user inputs and may use them to improve their models, that proprietary material left Samsung's control. The company banned public generative AI on its devices and networks and accelerated work on its own in-house model.
That episode is exactly the risk the new rollout is engineered around. Public chatbots, used casually, offer no data-processing agreement, no guaranteed data residency, and no reliable way to delete what was submitted — which is why simply telling employees to be careful does not work.
How the Security Gate and Two-Track Model Work
Samsung's answer has two technical pillars. The first is an access-control system: from April to May, Samsung Electronics' DX division ran a two-month proof-of-concept with 2,500 employees testing ChatGPT, Gemini, and Claude, and built a control system that grants tool access only to staff who complete internal security training. In enterprise practice, this kind of gating pairs identity-based access with data-loss-prevention controls that inspect prompts and block sensitive material before it reaches an external model — the interception layer that did not exist in 2023.
The second pillar is the contract tier. Enterprise versions of these tools differ from the public apps in a way that matters here: vendors including OpenAI, Google, and Anthropic commit, by default, not to train their models on enterprise customer data, and offer retention controls and data-processing agreements that the consumer versions do not. Running those enterprise tools alongside Samsung's in-house model, Samsung Gauss, gives the company a two-track approach: sensitive or proprietary work can stay on the internal model, while general productivity tasks can use whichever external model performs best. That structure also avoids locking the entire group into a single vendor.
Leadership Trains First, Then 2,300 Executives
Samsung is sequencing the rollout from the top. On the premise that a CEO's AI literacy determines the success or failure of the effort, the company will run an intensive "AX Boot Camp" — its first AI training aimed at the heads of affiliates — for about 50 presidents over two days this month at the Hoam Hall of its Human Resources Development Institute. Rather than passively attending, the presidents are expected to present their own proposals for redesigning each company's workflows with AI, and to jointly declare an "AX Vision."
Roughly 2,300 executives are then scheduled to train in three-day, two-night sessions through August 12, with training for all employees to be completed within the year. To sustain the effort, Samsung plans to set up dedicated AI organizations at every affiliate to craft business-specific strategies, manage data and model operations, and develop AI talent. CEOs are expected to lead the change by applying AI to eight core processes: development, procurement, manufacturing, logistics, marketing, sales, service, and management support.
Frequently Asked Questions
Did Samsung ban ChatGPT?
Yes. In 2023, Samsung banned employees from using public generative AI tools on company devices after engineers leaked confidential source code and a meeting transcript into ChatGPT. The new 2026 rollout re-admits these tools, but only through enterprise versions and behind mandatory security training.
What is Samsung Gauss?
Samsung Gauss is Samsung's in-house generative AI model. In the new rollout it runs alongside external tools in a two-track approach, letting Samsung keep sensitive work on its internal model while using outside models such as ChatGPT, Gemini, and Claude for general tasks.
Which external AI tools is Samsung adopting?
Samsung is rolling out Google's Gemini, OpenAI's ChatGPT, and Anthropic's Claude across all affiliates this month. It is the first time the conglomerate has adopted outside AI models companywide, after a two-month proof-of-concept with 2,500 employees.
How is Samsung keeping company data secure with these tools?
Access is granted only to employees who complete internal security training, and the tools are deployed as enterprise versions whose vendors commit not to train on customer data by default. Sensitive work can also be routed to the in-house Samsung Gauss model rather than an external one.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
