TikTok’s US Reset

How the US is asserting control over algorithms without banning foreign platforms

TikTok’s new US joint venture is being framed as a political compromise. In reality, it is something more consequential: a template for how the United States intends to govern foreign AI platforms without banning them outright.

This is not a forced sale. It is not a reprieve. It is localisation by design. Control over data, algorithms, and governance now matters more than who ultimately owns the shares.

Under the agreement finalised on January 22, TikTok’s US business now operates as an 80.1% American-owned entity, valued at approximately $14 billion. Oracle, Silver Lake, and MGX each hold 15% stakes as managing investors, with ByteDance retaining 19.9%. The new entity has its own CEO, security leadership, and independently governed board.

ByteDance keeps economic exposure to the platform’s estimated $11.8 billion in US advertising revenue projected for 2025. However, the most sensitive functions data access, algorithm development, and oversight are now anchored in the United States. This is not divestment. It is structural separation.

The most important change is not ownership, but the algorithm carve-out. TikTok’s US entity will retrain, test, and update its recommendation system using only US user data, under US governance and security controls. TikTok’s algorithm is its moat. Allowing it to be split in practice creates two systems instead of one global engine. One evolves under US rules. The other serves the rest of the world.

That breaks a long-standing assumption in platform design that large AI systems must operate with a single, unified intelligence core. What TikTok has accepted is a divided system. The US feed will now change based on US legal, political, and cultural pressures. This accelerates a broader shift toward region-specific AI, something regulators in Europe and India have discussed for years, but which TikTok has now implemented at scale across 200 million US users and 7.5 million US businesses.

A critical beneficiary of this structure is Oracle. Oracle is not simply providing cloud infrastructure. It is serving as the trusted intermediary that regulators can see and audit. US user data stays on US servers. Access is logged. Oversight is built into the system. Oracle’s role makes the arrangement credible in Washington.

We have seen this pattern before. Microsoft plays a similar role for OpenAI, combining deep infrastructure dependency with long-term commercial and intellectual property agreements. Amazon does the same for government computing. As platforms increasingly sit at the intersection of media, AI, and public influence, trusted infrastructure is no longer optional. It is the cost of operating at scale.

The board structure reinforces that message. With representation from Silver Lake, TPG, Susquehanna, Oracle, DXC, and MGX, this entity looks less like a consumer tech company and more like a piece of national infrastructure. Creative discretion has been narrowed in favour of institutional oversight. The signal from Washington is clear. When national security is involved, stability matters more than vision.

The political framing matters too. By publicly endorsing the deal, the administration effectively reclassified TikTok. It moved from being treated as a foreign risk to being recognised as a domesticated platform. That makes future bans harder to justify and provides protection during an election cycle. China’s restrained response suggests this was the least damaging outcome available. A full forced sale would likely have crossed a red line. What emerges is managed compromise, not surrender on either side.

For other platforms, this agreement reads like a playbook. Any foreign-owned, AI-driven company operating at national scale, especially one that shapes public attention or discourse, should expect similar demands. Local data storage. Local algorithm training. Independent national boards. Security leadership with government credibility. Global reach through a single system is no longer the default.

For media companies and advertisers, this reframes platform risk. TikTok US now looks less like an unpredictable social app and more like a regulated broadcaster with defined guardrails. That likely improves brand safety and advertiser confidence. The trade-off is fragmentation. Campaigns, optimisation, and audience behaviour may increasingly differ by geography rather than converge globally.

For the 7.5 million US creators whose livelihoods depend on the platform, this divergence matters. Algorithmic separation means content may surface differently in the United States than elsewhere, reshaping discoverability, reach, and monetisation patterns over time.

The deeper signal goes beyond TikTok. This move follows the same logic already visible in strategic technology controls. Huawei was constrained not by demand, but by access to advanced chips. ARM sits under national scrutiny because its designs shape the processor ecosystem. ASML is restricted because its machines determine who can manufacture leading-edge semiconductors.

Different industries. Same approach.

In each case, the state did not seize assets outright. It asserted control over the most important layer.

TikTok shows that algorithms now belong in that category. Recommendation systems, training pipelines, and data feedback loops are being treated as strategic infrastructure rather than neutral software. Control, not ownership, is the key issue.

This marks the end of borderless AI as a guiding assumption. AI systems will be shaped by national boundaries. Trust and oversight will matter as much as technical performance. Platforms that shape culture, media, and political attention will increasingly face a choice between adapting locally or being shut out.

TikTok did not just save its US business. It accepted that algorithms, like chips and networks before them, are now governed as instruments of state power.

-Maureen

Image created by OpenAI