Samsung Reverses Its ChatGPT Ban - and Lands One of OpenAI's Largest Enterprise Deals Ever

Three years ago, Samsung Electronics made global headlines by banning ChatGPT and every other generative AI tool across the company after engineers accidentally leaked sensitive source code through the chatbot. On June 22, 2026, the company completed a remarkable about-face: it is now deploying ChatGPT Enterprise and Codex to tens of thousands of employees worldwide in what OpenAI explicitly describes as "one of OpenAI's largest enterprise launches ever" 1.

What Was Announced

The deal covers two distinct employee populations. ChatGPT Enterprise and Codex are now available to all Samsung Electronics employees in South Korea, and to every employee worldwide in its Device eXperience (DX) Division - the business unit responsible for Galaxy smartphones, tablets, and home appliances 1. The deployment is not a limited pilot; it is being rolled out as a core operational platform across the conglomerate 2.

Samsung plans to embed both tools broadly. The company intends to use ChatGPT and Codex across software development, marketing, product development, and manufacturing 1. On the knowledge-work side, ChatGPT handles research, drafting, data interpretation, and documentation 1. Codex - OpenAI's coding agent - lets employees turn ideas into working software, internal tools, websites, and automated workflows, even without a coding background 1.

Notably, Samsung is deploying ChatGPT Enterprise alongside Google's Gemini and Anthropic's Claude as part of a multi-vendor AI stack, giving employees flexibility depending on their task requirements 3.

The Numbers Behind the Deal

The scale of adoption that preceded the full rollout is striking. According to OpenAI's official announcement, more than 5 million people now use Codex every week for technical and non-technical workflows 1. Codex weekly active users in South Korea alone surged nearly 800% between February 1, 2026 and the announcement date - well before the enterprise-wide contract was signed 1.

OpenAI and Samsung have history beyond software. In October 2025, OpenAI, Samsung Electronics, Samsung SDS, Samsung C&T, and Samsung Heavy Industries signed a letter of intent covering global AI data center infrastructure, with Samsung Electronics designated as a strategic memory partner to supply advanced semiconductor solutions for OpenAI's Stargate initiative 4. The new ChatGPT Enterprise agreement expands that infrastructure partnership into enterprise workforce transformation 1.

How Samsung Got Here - and Why It Matters

The backstory is essential context. In March 2023, Samsung engineers accidentally uploaded sensitive source code and internal meeting notes to ChatGPT. Samsung's response was swift: a company-wide ban on generative AI tools 2. The reversal began taking shape in late 2025, when Samsung SDS - the group's IT services arm - became the first Korean company to offer ChatGPT Enterprise and provide technical support to enterprise customers in Korea, signing a reseller partnership agreement with OpenAI on December 23, 2025 5.

Before committing to OpenAI at scale, Samsung ran a competitive evaluation. From April to May 2026, Samsung's DX Division conducted a two-month proof-of-concept with approximately 2,500 employees, testing ChatGPT, Gemini, and Claude and evaluating security controls and operational efficiency before proceeding with the full rollout 67. Access to the tools is gated: employees must complete an internal corporate AI security compliance training course before receiving active credentials 6. Under the enterprise architecture, no data, prompts, or code entered by Samsung employees can be used to train OpenAI's models, and all data remains sandboxed within a managed corporate cloud environment 8.

Samsung's semiconductor Device Solutions division - home to some of the company's most sensitive chip IP - reportedly still faces tighter restrictions on external AI tools 9. Full global workforce training is expected to be completed by the end of 2026 2.

Samsung SDS's reseller arrangement also gives OpenAI a new institutional distribution channel into Korea's broader enterprise market. A growing number of Korean companies - including LG Electronics, LG Uplus, LG CNS, GS E&C, TVING, Krafton, Toss, MUSINSA, Korea Zinc, Nexen Tire, and HanaTour - are also using ChatGPT Enterprise, OpenAI APIs, and Codex 1.

What This Signals for Enterprise AI Adoption

The Samsung win is significant for at least three reasons.

It validates the enterprise-grade security story. The 2023 leak made Samsung a cautionary tale cited in virtually every enterprise AI risk discussion. The fact that the same company has now deployed ChatGPT at scale - after a formal multi-vendor evaluation and a structured security-gating process - is a powerful data point for other large organizations still on the fence.

It demonstrates that Codex has crossed from developer tool to company-wide platform. Samsung is deploying Codex not just to its engineers but to marketing, manufacturing, and product teams. The combination of ChatGPT for knowledge work and Codex for automation means Samsung is treating AI assistance as a unified capability rather than a set of departmental point solutions 1.

It intensifies enterprise competition among the frontier labs. OpenAI explicitly ranks the Samsung deal among its largest enterprise launches ever 1. For Anthropic and Google - both of which competed in Samsung's DX Division evaluation - this contract going to OpenAI is a concrete signal of how enterprise procurement decisions are breaking at the top of the market.

Why It Matters

Samsung's trajectory - from a 2023 ban to a 2026 flagship deployment - is a compressed version of the arc every large organization is navigating right now: how to capture AI productivity gains without creating new data-security liabilities. The answer Samsung landed on involves mandatory training gates, enterprise-grade data isolation, a multi-vendor evaluation, and a phased rollout. That playbook, more than the contract itself, may be the most replicable lesson for the thousands of enterprises still working out their own AI policies.