Researcher Restriction Index.
China's senior AI researchers are increasingly being kept on Chinese soil. Since May 2026, passport custody and pre-travel-approval restrictions — until then concentrated in state labs — have expanded to senior researchers at Alibaba, Moonshot, Zhipu, and DeepSeek. This paper builds a forward-looking risk index that scores each Chinese AI researcher's chance of facing a travel curb, lab freeze, or formal restriction in the next twelve months. We publish the score on a small cohort (five pseudonymized profiles), show every underlying feature and line of code, and grade the model against the historical record. The model is right about 92% of the time on this small set; two of the five profiles now sit at the 0.80 published-score ceiling. Each score will be marked right or wrong twelve months from its publication date.
What we've actually found.
U.S. co-authorship flipped from protective to risk-loading in Oct 2022.
Before the BIS chip export controls, Chinese AI researchers with dense U.S. co-author ties were less likely to face restrictions — Beijing wanted to retain them. After the controls, the same ties became a vulnerability.
Two of five tracked researchers sit at our 80% confidence ceiling.
We cap published scores at 0.80 by editorial guardrail — anything more confident than that requires a larger cohort. The fact that two profiles hit the ceiling already says the post-2022 risk regime is structurally tight.
Travel curbs on senior AI researchers expanded from state labs to private firms.
Since May 2026, Bloomberg has reported that passport custody and pre-travel approval — until then a state-lab practice — now apply at Alibaba, Moonshot, Zhipu, and DeepSeek. This expansion is the index's founding event.
The full historical-event corpus, the per-profile scoreboard with confidence intervals, the feature definitions, the held-out backtest with three baselines, the editorial guardrails, and the open questions all live inside the v0 founding paper.
Published pieces.
Each card is a standalone read — a substantive finding or a release. Methodology hygiene, propagation fixes, and audit notes live in the Changelog below.
Before October 7, U.S. ties protected Chinese AI researchers. Now they don’t.
We taught the model to see the October 2022 U.S. chip-export controls as a turning point in how Beijing treats its own AI researchers — and the data now shows the flip clearly.
U.S. co-authorship used to insulate Chinese AI researchers from Beijing’s scrutiny. After the chip controls, the same ties became a liability. Our predictions get materially better once we let the model see the break.
Introducing the Researcher Restriction Index.
The founding paper of the index. We forecast each Chinese AI researcher’s chance of facing a travel restriction in the next twelve months — with every input, every assumption, and the source code itself shown openly.
Public, dated forecasts on real people that get graded twelve months later. Most China analysis stays inside think tanks; this one is published openly and held accountable to the record.
Methodology, propagation, audits.
Smaller fixes — primary-source audits, page/code reconciliation, engineering — that aren't publishable on their own but are recorded here for traceability.
- 31 May 2026Details →Page and Python reconciledThree corrections logged in the prior pass hadn't all landed in the Python pipeline. Closed the gap so the pipeline can re-run without reverting published findings.
- 31 May 2026Details →Primary-source pinning + cohort expansionAudit caught a fabricated Bloomberg URL, a misattributed dispatch, and one over-extrapolated firm in the inductive corpus. The public-source-only guardrail worked.