With Google and ByteDance (字节跳动) Holding All the Cards, Why is Programming Still Their Weak Point?

What Samsung will change

It has been reported that Samsung Electronics will begin allowing employees to use third‑party generative AI models from next month, ending a period in which staff were limited to the company’s in‑house model, Samsung Gauss. Reportedly the rollout will first target the Device eXperience (DX) division — the group that handles displays, mobile devices and home appliances — with access opening sometime in June. Company sources say employees will need to complete mandatory security training before they can use external models; the semiconductor arm, Device Solutions (DS), will remain tightly restricted.

Details and internal training

It has been reported that Samsung plans an on‑site AI training programme later this year aimed at roughly 2,000 executives from Samsung and its major subsidiaries, intended to teach safe use of AI applications. The change follows internal memoranda flagged by the Korea Times and relayed by other outlets. Why the shift now? Productivity gains are obvious. But security and intellectual‑property risks are not trivial, especially where chip design and supply‑chain secrets are involved.

Why it matters beyond Samsung

For Western readers: Samsung is South Korea’s largest technology conglomerate and a major global supplier of chips, phones and displays. Its cautious opening underscores a wider industry problem. Google and ByteDance (字节跳动) — among others — control many of the most advanced generative models and developer tools. Yet programming and code‑generation remain a weak spot for many vendors trying to internalise AI capabilities. Can large device makers build rivals to the cloud‑native models that dominate today? Or will they increasingly rely on external providers while wrestling with security, export controls and geopolitical friction?

The bigger picture

Reporters note this is not just a product decision but a policy one: companies must balance competitiveness against national security concerns and tightening export rules on semiconductors and AI tooling. Samsung’s phased, department‑specific approach is cautious by design. It signals that even firms with deep R&D pockets are finding it easier to tap external AI — but harder to trust it with their most sensitive work.