"Humans-at-the-Helm" - Not "Humans-in-the-Loop"

(Proposed Non-Agentic Public Infrastructure for All Government Levels)

Can democratic institutions gain better digital sight without giving AI authority?

Two new Civic AI thought experiments consider how public institutions might detect and preserve early signs of digital disturbances while keeping judgment, responsibility, and operational authority human.

I recently joined the Apolitical community and would welcome the perspectives of public servants, policymakers, civic technologists, cybersecurity professionals, and public-administration colleagues on a question that may become increasingly important as governments confront more frequent, disruptive, and costly digital pressures.

Public institutions already operate amid difficult conditions: cyberattacks, ransomware, data extortion, manipulated online narratives, strained public-facing systems, limited staffing, complex legal responsibilities, and public expectations for timely information and response.

AI may be able to help organize records, preserve provenance, identify aggregate changes, translate authorized information into usable public formats, and make developing conditions more visible.

But an earlier question must come first. Can a democracy use advanced AI to see a changing civic digital field more clearly without allowing that system to become an investigator, intelligence actor, recommender, incident commander, or governing authority?

I explore that question in my new theoretical proposal, U.S. Civic AI Course Correction: Humans-at-the-Helm - Not Humans-in-the-Loop (Civic Digital Bifocal Resolution for Visibility Without Authority - Non-Agentic Civic AI in Foreign-Origin and Domestic Stress Tests).

The work is not an operational cyber-defense plan or a claim that AI can solve a public crisis. Instead, it tests a narrower proposition: a Civic AI system may continuously preserve, organize, measure, translate, and illuminate public conditions without acquiring authority to interpret, attribute, recommend, command, or decide.

The paper calls this approach Non-Agentic Civic AI, defined as civic information infrastructure designed to provide visibility without authority.

It examines two contrasting stress tests.

The first, at the External Edge, considers fabricated infrastructure-failure narratives spreading during an election week. The question is whether a system could render lawfully observable foreign-origin digital patterns, preserve evidence and uncertainty, and make the field visible - without identifying actors, inferring intent, attributing responsibility, controlling content, or entering a national-security command chain.

The second, in the Domestic Core, considers a municipal ransomware and data-extortion disruption affecting public-facing systems. Here the question is whether a system could preserve public notices, record service disruptions, display aggregate conditions, and maintain a transparent civic record - without becoming a domestic investigator, incident commander, ransom negotiator, legal authority, or operational response system.

The architecture proposes several deliberately bounded functions:

• A civic-memory function that preserves authorized public records, timestamps, provenance, and document history.
• An aggregate-conditions function that can show changes in public-service disruption without profiling citizens or assigning risk scores.
• A translation function that standardizes authorized information for human use without prioritizing, recommending, or deciding.
• A continuously revised public Civic Digital Visibility Record distinguishing among what is observed, corroborated, pattern-indicated, unresolved, and requiring human inquiry.
• A permanent disclosure of methodological limits, including what the system cannot see, establish, authenticate, or lawfully disclose.

The central claim is not that machine systems should become passive. It is that they can remain active in visibility while remaining inactive in authority.

A system may provide more continuity, resolution, evidence preservation, historical context, and public legibility during a crisis. Yet attribution, legal interpretation, prioritization, proportionality, negotiation, remediation, and accountability should remain with constitutionally authorized human institutions.

In other words, the machine may help resolve the civic digital field, but human institutions must resolve the crisis.

I would welcome the practical experience and concerns of those working in government and public policy:

• Where could continuous, aggregate civic visibility be genuinely useful during digital disruption?
• What safeguards would prevent a visibility system from gradually becoming a recommendation, targeting, surveillance, or command system?
• How can public institutions make uncertainty and methodological limits visible rather than allowing incomplete information to become premature machine certainty?
• What would make an institutional handoff to human officials meaningful rather than merely formal?
• Could a system designed to preserve records and show aggregate conditions improve democratic accountability without entering the chain of consequential public decisions?

The complete piece can be quickly accessed gratis, without any website registration, here:

medium.com/@crob2011/u-s-civic-ai-course-correction-humans-at-the-helm-not-humans-in-the-loop-0608c584de0f

You can also access it through a browser search of my full name.

The argument is deliberately ambitious but also bounded. Democratic institutions may be able to give themselves better sight without giving a machine authority over what that sight means - or what must be done next.