The company stayed. The business heart moved.

Traditional acquisitions have visible boundaries: shares, business units, factories or IP move. In AI, a different pattern has emerged — the legal entity stays while founders, core researchers, engineers and technology access move together.

That creates a harder question. If the buyer never acquired the company, is this merely hiring? Or, if the incoming team enables the buyer to continue the same business activity, has part of the business effectively moved?

The KFTC’s proposed amendment announced on September 9, 2026 addresses that boundary and explicitly discusses acqui-hire transactions in AI and other advanced-tech sectors.

The question is not how many people were hired

The proposed guidance is not a headcount rule. It looks at whether organized personnel and the technology and know-how they hold perform a core business function, and whether their transfer allows the acquirer to carry on the existing business activity.

Ordinary job moves or recruiting a few employees from a competitor do not automatically become mergers. The key question is whether the organized transfer effectively moves a major part of the business.

That is the difference between an HR event — key people left — and a competition-law event — business capability moved.

An AI company’s technology is not stored only in code

Code, models, data and patents matter, but knowledge about failed data, misleading evaluations, abandoned architectures and field failures often sits in team experience.

That knowledge can be distributed across researchers, platform engineers, product leaders and executives. Part of the technology therefore lives in the connections among people — in shared memory built through repeated execution.

This is why moving an intact team can have a different economic meaning from hiring the same number of strong individuals separately.

Microsoft–Inflection AI and Google–Character.AI showed the new boundary

In 2024 Microsoft hired Inflection AI co-founder Mustafa Suleyman and several other employees while also securing access to Inflection’s AI models. Reuters reported payments of about $650 million around the arrangement. The UK CMA opened a formal investigation into whether the hiring and partnership amounted to a merger.

Also in 2024, Character.AI co-founders Noam Shazeer and Daniel De Freitas returned to Google, with some colleagues moving as well. Google simultaneously obtained a non-exclusive license to Character.AI’s model technology.

The entities remained, yet people and technology rights moved together. That is why acquiring a capability can sometimes describe the transaction better than acquiring a company.

The minimum unit of acquisition may shrink from a company to a team

Reading this merely as an intensifying AI talent war misses the deeper shift. The unit of acquisition itself can become smaller.

If a small team combines research know-how, product judgment, data practice and deployment experience into a core operating function, that team may be closer to the economic object of the transaction than the whole legal entity.

Company Acquisition begins to decompose into Capability Acquisition — especially where a small group carries a disproportionate amount of tacit knowledge.

Talent management may need to map connections, not just individuals

Companies usually identify key talent one person at a time. The acqui-hire lens suggests that individual lists may miss the more valuable asset.

Hiring ten excellent people separately is not the same as moving ten people who have spent years building the same product together. The latter carries shared memory about who knows what, which judgments are trusted, and which paths have already failed.

That is more than teamwork. Economically, it can be an asset that compresses learning time.

Banseog View — the scarce AI asset may be a capability that can move together, not just one star hire

The striking part of the KFTC proposal is that it views people movement in the language of business functions, not HR. The question becomes not who moved, but what capability appeared at the receiving company because they moved.

That lens also changes talent risk management. A company needs to know not only who is important, but which capabilities disappear when certain combinations of people leave together.

In AI, talent is not merely an input to the business. In some cases, the organization of talent can itself be part of the business.

Primary sources and references

The proposed KFTC guidance does not mean ordinary employee moves or routine recruiting are mergers. The analysis depends on whether organized key personnel plus technology and know-how transfer a substantial business function, together with applicable statutory notification thresholds. The proposal remains under public consultation through September 30, 2026.