When "Data for Good" Is Just Surveillance in a Lab Coat
Talking Points Memo has reported that "the National Institutes of Health (NIH) appears to be repurposing a project that Congress shot down last year."
Congress already killed this project once.
It was called an Alzheimer’s "real-world data platform" — a $300 million plan to scrape personal information from phone providers, wearable devices, insurers, EHRs and government agencies. GAO found it riddled with management failures and privacy blind spots. It died because the risks outweighed the promise.
Now NIH has brought it back from the dead, rebadged for autism, with Robert F. Kennedy Jr. promising to find "the cause" by deadline and Jay Bhattacharya’s NIH obliging with a four-month sprint. The new name is Autism Data Science Initiative. The methods and contractors — Palantir among them — largely overlap with the Alzheimer’s project. So do the structural problems Congress flagged.
HHS says it’s not a "registry." Officially, it isn’t. Functionally, if you aggregate, persist and link records across agencies to track a named population, you’ve built one — whatever you call it. And registries come with a history: they outlive their stated purpose, migrate into other hands and get used in ways participants never consented to.
This is not a paranoid hypothetical. The Alzheimer’s version failed to safeguard privacy, failed to limit scope and failed to prove it could deliver insight without collateral harm. Congress urged NIH to pause. Instead, the agency swapped the disease area, kept the architecture and tightened the timeline. That is not science. In effect, it is laundering surveillance through a different diagnosis.
Autistic people are not raw material for a Cabinet secretary’s pet theory — or for any political vanity project. We are not datasets to be mined until someone’s deadline arrives. And we know what happens when you centralize "de-identified" health data: it gets re-identified. It gets shared. It gets subpoenaed.
If NIH wants credibility, here’s the minimum:
- Consent that is opt-in, revocable and binding downstream — no silent inclusion, no retroactive justifications
- Purpose limits with teeth — no law enforcement backdoor, no benefits-eligibility fishing, no "public health" mission creep
- Autistic governance with veto power — not an advisory role, actual authority
- Sunset clauses and independent audits — because a platform without an end date is an infrastructure for control
Until then, this is not "data for good." It is another system asking autistic people to fund our own precarity with our medical exhaust.
Congress has already shown bipartisan discomfort. That is our opening. If your senator sits on Appropriations, tell them to hold NIH to its own "pause" language before this reboot becomes the default architecture for tracking us all.