Sources: seeking to win US allies, TikTok has proposed a complex, $1.5B plan to reorganize its US operations, including oversight of its key algorithms
Chinese-owned app proposes giving U.S. officials oversight of its algorithms — Two years into negotiations with U.S. regulators …
Context & Ripple Effects
This $1.5B proposal is the latest move in a negotiation that has dragged since at least 2020, when [[a:957446|China's new AI export restrictions left sale parties unsure whether TikTok's algorithms could even change hands]]. In December, TikTok tried appeasing Washington by offering to operate more of its business at arm's length with outside scrutiny from Oracle and others, and it has paired the concessions with a transparency charm offensive that echoes Huawei's failed playbook in Europe.
What is new here is the target: not data storage but the recommendation engine itself, with US officials getting oversight of the key algorithms. That is the piece Beijing's export rules treat as strategic property, which is why the plan is structured as oversight of a Chinese-owned asset rather than a transfer.
First-order effects
- US officials would gain a formal window into how TikTok's For You feed is built and tuned, while TikTok keeps ownership of the algorithm — a middle path between full divestiture and the status quo.
- Oracle's role as trusted third-party scrutineer deepens, extending the monitoring arrangement TikTok already offered into the algorithm layer itself.
Second-order effects
- Skeptics inside the government argue the oversight scheme faces practical difficulties and may not settle the core question of possible Chinese influence, keeping pressure alive for harsher remedies such as forced divestiture.
- Beijing's position becomes the binding constraint: related reporting details Chinese-government tensions over any deal, including a possible veto, meaning the plan's fate depends as much on export-control politics as on US acceptance.
Third-order effects
- If oversight-of-a-copy becomes the accepted template, the endgame looks like the structure the White House later described — a US entity leasing a copy of ByteDance's algorithm for Oracle to retrain — establishing state-mediated algorithm access as the standard settlement for foreign-owned AI platforms.
- The Huawei parallel cuts the other way: if transparency tours and technical concessions fail to buy political acceptance, Western governments may conclude that only ownership separation resolves national-security concerns about recommendation systems.
The trend: Cross-border AI platforms are converging on state-mediated algorithm access — leases, copies, and third-party retraining — as the compromise between national-security demands and China's export controls.