Three years ago Samsung banned the use of ChatGPT after sensitive information was uploaded to external servers. Today, the South Korean giant has completely reversed course: employees of Samsung Electronics in Korea and the global Device eXperience division are getting access to ChatGPT Enterprise and Codex, AI tools for technical and non-technical work, from software development to marketing and manufacturing.

From fears to controls: the Enterprise formula

The large-scale return is made possible by the enterprise version of OpenAI's chatbot. Unlike consumer products, ChatGPT Enterprise offers granular controls for data protection, access management and user administration. Organisations can apply internal security policies, decide who can use the tool and monitor information flows. OpenAI highlights that the deployment covers all Samsung Electronics employees in Korea and all DX division staff – the unit that includes smartphones, consumer electronics and home appliances – without being limited to specific business units.

The announcement marks Samsung's intention to move past the concerns that led to the ban in 2023: generative AI is now entering company processes with the promise of supporting tasks such as information research, document drafting, idea development, data interpretation and code production.

Codex: code and much more

Alongside the chatbot, Codex enters the scene, the tool aimed at writing, reviewing and debugging code. OpenAI reports over 5 million weekly users and, for Korea, an almost 800% growth in attention since the start of February 2026. Codex will not only serve developers: Samsung will also use it to build internal tools, websites, software prototypes and automated workflows, involving non-technical teams in creating digital solutions.

Harrison Kim, general manager of OpenAI Korea, describes it as one of the largest enterprise deployments ever signed, stressing that AI is distributed across functions rather than confined to specific departments.

The memory thread: Samsung and the Stargate supply

The decision does not stand alone. In October 2025 Samsung had announced a partnership as a strategic memory supplier for OpenAI's Stargate infrastructure initiative. OpenAI's memory demand, according to South Korean government sources cited by Reuters, could reach 900,000 DRAM wafers per month. Samsung Electronics and SK Hynix together control about 70% of the global DRAM market and nearly 80% of the high-bandwidth memory (HBM) market, essential for fast data movement between memory and processors in AI systems.

Samsung SDS, the IT services subsidiary, has also started collaborations with OpenAI to jointly develop AI data centers and provide consulting, deployment and management services for companies integrating OpenAI models into their internal systems. A reseller agreement will also allow Korean companies to adopt ChatGPT Enterprise through Samsung SDS.

The deployment knot: controlled cloud or on-premise?

For those watching the enterprise AI market, Samsung's move is exemplary: after initial fears about data sovereignty, the answer was not abandonment but adoption of a cloud platform with strengthened security controls. This is an increasingly common path, as shown by Deloitte's 2026 findings: 66% of organisations report productivity or efficiency gains, 53% improvements in analysis and decision-making.

Yet the conversation doesn't end there. The same study and a Bpifrance survey of French mid-sized companies show that only 17% of firms using generative AI see tangible time savings. Simple availability of tools is not enough: deep process integration, training and a clear total cost of ownership (TCO) evaluation are needed. Moreover, for organisations with strict data residency obligations or sector regulations, cloud access – even with enterprise-grade features – may not be sufficient. The debate over cloud versus on-premise deployment remains central, and Samsung's choice shows how even industrial giants seek compromises between adoption speed and governance controls.

Samsung's opening to ChatGPT Enterprise and the simultaneous push on the memory supply chain confirm a trend: generative AI is becoming a pervasive instrument, with implications that go far beyond writing text or code, reaching into the hardware supply chain and data security strategy.