Samsung Rolled Out ChatGPT Enterprise Three Years After Banning It

On June 21, Samsung Electronics deployed ChatGPT Enterprise and Codex to every employee in South Korea and every worker in its global Device eXperience division — one of OpenAI's largest enterprise rollouts ever, announced three years after the same company banned generative AI when engineers pasted proprietary source code and confidential meeting notes into the public version of ChatGPT.

The reversal: In May 2023, Samsung blocked ChatGPT, Google Bard, and Bing on company devices after Bloomberg reported that staff had uploaded sensitive internal code to OpenAI's platform. The Korea Herald later detailed three separate leaks in the semiconductor division — including source code and meeting minutes fed into ChatGPT within weeks of initial access. Samsung's internal memo warned that failure to comply could mean "disciplinary action up to and including termination of employment." The ban became the cautionary tale every CISO cited.

Why now: Samsung did not soften its stance because models got smarter. It hardened the contract. ChatGPT Enterprise commits by default not to train on customer data, with identity-based access controls and data-processing agreements the consumer product never offered. Samsung ran a two-month pilot with roughly 2,500 DX employees testing ChatGPT, Gemini Enterprise, and Claude simultaneously before the June 21 formal agreement with OpenAI, per CIO and Dataconomy. Access requires mandatory internal AI security training — the control layer Samsung built specifically to avoid a repeat.

The supplier-customer knot: Samsung is not just adopting OpenAI's software. It is supplying the memory OpenAI's infrastructure runs on. Chairman Lee Jae-yong and CEO Sam Altman signed a letter of intent in October 2025 for Samsung to provide advanced semiconductors for OpenAI's Stargate data center buildout, The Korea Times reported. The June 21 deal expands the relationship from silicon supplier to workforce platform — the same week SK Hynix overtook Samsung as South Korea's most valuable listed company on pure memory exposure.

Not a monoculture bet: Samsung's DX division opened ChatGPT, Gemini Enterprise, and Claude to employees on June 12 as part of its "AX" AI transformation push, CIO reported. Samsung still runs its in-house Gauss model for sensitive tasks. The strategy is vendor pluralism with enterprise guardrails — not blind faith in one lab.

Codex goes beyond developers: The June 21 agreement formalizes Codex alongside ChatGPT Enterprise — OpenAI's agentic coding tool now pitched to marketers, product teams, and manufacturing staff building internal tools and automated workflows. OpenAI says more than 5 million people use Codex weekly, with Korea's weekly active users up nearly 800% since February 1. Harrison Kim, general manager of OpenAI Korea, called the deployment "historic" because Samsung is "embracing AI not as a tool limited to certain teams or functions, but as a core platform for improving how employees around the world work and innovate."

The IPO backdrop: OpenAI reportedly filed confidential S-1 paperwork in May and is targeting a public debut as early as Q4 2026, per TechTimes and other outlets. Enterprise revenue is the growth story investors will underwrite — and Samsung is the reference customer that proves a company that once fired ChatGPT for leaking secrets will now bet its global DX workforce on it. LG Electronics, Krafton, Toss, and Seoul National University's 47,000-person ChatGPT Edu rollout show Korea is becoming OpenAI's enterprise beachhead in Asia.

What to watch:

• Whether Samsung's Device Solutions semiconductor division expands beyond ChatGPT access to full Codex deployment
• How Samsung's multi-vendor AX strategy holds when labs compete for the same workflows
• Whether enterprise contract terms survive the first major incident — Samsung's 2023 ban followed three leaks in the semiconductor division within weeks
• OpenAI's S-1 disclosure of Samsung-scale deployments as enterprise revenue proof points

Convina's view: Samsung's ban was never an anti-AI statement. It was a procurement failure — consumer ChatGPT had no data boundary, and employees treated it like a search bar. The June 21 deal is the market pricing the fix: enterprise contracts, security training, and vendor pluralism as the real product. That is good news for OpenAI's IPO narrative and a warning for every lab still selling intelligence without liability architecture. The company that supplies OpenAI's memory just validated OpenAI's enterprise stack with its own workforce. If that does not clear the bar for your board's AI rollout, nothing will — and if it does, the question is no longer whether to deploy, but which vendor's contract you trust when the leak happens anyway.