Civic
A new CDT survey reveals deep, bipartisan anxiety about how public agencies collect, store, and use personal information.
NewsOnScale Staff
October 22, 2026
The timing is difficult to ignore. As federal and state agencies race to integrate artificial intelligence into everything from benefits administration to law enforcement, a new survey from the Center for Democracy and Technology finds that a majority of Americans are not just skeptical of how government handles their personal data — they are actively worried about it.
The CDT survey, released this week, documents widespread concern about data collected by public agencies: who can access it, how long it is retained, whether it is shared with third parties, and what recourse ordinary citizens have when something goes wrong. The findings are not a partisan story. Anxiety about government data practices cuts across Republican, Democrat, and independent respondents, suggesting this is less a political flashpoint than a structural crisis of institutional trust.
## What People Are Actually Afraid Of
The survey does not describe a population that is abstractly worried about privacy in some theoretical sense. Respondents pointed to specific, concrete fears: data being used for purposes beyond what it was originally collected for, information being shared with law enforcement or immigration authorities without notice, and a near-total absence of meaningful mechanisms for citizens to see or correct what agencies hold about them.
These are not irrational fears. In recent years, documented cases have emerged of social services databases being cross-referenced with immigration enforcement systems, of predictive policing tools trained on historically biased arrest records, and of benefits algorithms flagging vulnerable people for fraud investigations with no human review and no clear appeals process. The public's instincts are tracking a real pattern.
## The AI Acceleration Problem
What makes this moment distinct — and what connects this survey directly to the broader AI agent economy — is the speed at which automated systems are now being layered onto existing government data infrastructure. Agencies adopting AI tools to process applications, flag anomalies, or route cases are, in effect, amplifying the reach and consequence of whatever data they already hold. An error in a legacy database that once inconvenienced one person can now, through an AI decision system, propagate that error across an entire case file, a benefits eligibility determination, or a background check.
The accountability gap here is not hypothetical. Most AI procurement by government agencies happens with minimal public disclosure. Contracts with vendors are frequently shielded from full transparency under trade secret exemptions. And the agencies themselves often lack the internal technical capacity to audit the systems they have purchased, let alone explain their outputs to affected citizens.
## Transparency Is Necessary But Not Sufficient
This is where the CDT findings intersect with a broader debate in civic technology. Posting a privacy policy, publishing a data inventory, or releasing a vendor contract — these are transparency measures, but they are not accountability. Accountability requires that someone with actual authority can act on what the disclosure reveals. It requires that citizens have standing to challenge decisions made about them. It requires that the cost of getting it wrong falls on the institution, not only on the person harmed.
Right now, that architecture largely does not exist for AI-assisted government decision-making in the United States. There is no federal framework governing algorithmic accountability in public agencies. State-level efforts are uneven and often outpaced by procurement cycles.
## What Comes Next
The CDT report calls for stronger data minimization requirements, clearer limits on secondary use of government-collected information, and enforceable rights for individuals to access and contest their records. These are reasonable, technically achievable asks. The barrier is political will and institutional inertia, not engineering.
What the survey makes undeniable is that the public is ahead of policymakers on this. Americans already understand, at a gut level, that data in government hands is not neutral. They want rules that match that reality. The question is whether the agencies now deploying AI at scale will move to earn that trust back — or continue to treat public concern as a communication problem rather than a governance one.