The Digital Governance and Artificial Intelligence Act
- Legislative Pathway: Reconciliation-Eligible (Major infrastructure and agency build-out; standalone for regulatory provisions)
- Goal: To ensure that Artificial Intelligence and digital infrastructure are used to enhance human agency, government efficiency, and economic equity — and to prevent the worst-case scenario where opaque algorithmic systems make consequential decisions about Americans’ lives without accountability, recourse, or public oversight.
The Digital Front Door (Open Data Architecture)
Mandate that all federal agencies migrate to a unified, open-source data architecture, enabling the “Digital Front Door” where citizens can access all services through a single, secure interface.
- Digital Sovereign ID: Optional federal identity framework that gives citizens a single secure credential to access healthcare, Job Guarantee, voting, tax filing, and benefits — without monopolistic private intermediaries (Login.gov expanded and hardened)
- Privacy by design: No central database of citizen activity; identity verification works through cryptographic attestation, not data aggregation
- Open source mandate: All government-built or government-funded software is open source by default (classified national-security systems excepted) — see Government Transparency Act for the parallel provision
- Interoperability standards: Federal data architecture published as standards; states and localities receive grants to adopt them
- Sunset on legacy systems: 5-year phase-out for COBOL and other unmaintainable legacy systems; modernization grants to agencies
AI for the People (Strategic Task Force)
Establish a cabinet-level National AI Council to manage the “Automation Transition” — using AI to make government work better for citizens while protecting workers and citizens from displacement and discrimination.
- Mandate to automate tedium: FOIA processing, permit approvals, IRS audit triage, Social Security benefit calculations, immigration paperwork — the high-volume tasks that frustrate citizens and burn out civil servants
- Workforce coordination: Direct coordination with Federal Job Guarantee (Economic Opportunity Act) for any private-sector displacement
- Domestic AI capacity: Federal investment in non-corporate AI infrastructure — public compute clusters available to researchers, public-interest projects, and small businesses
- AI Safety Institute formally chartered: Builds on the 2023 NIST AI Safety Institute with statutory authority, dedicated funding, and red-team capacity for evaluating frontier models
- No autonomous lethal-decision authority: AI systems may not make autonomous decisions about applying lethal force in any context
Algorithmic Transparency and Auditability
Any AI tool used by the government for consequential decisions must be auditable, explainable, and subject to human appeal.
- Covered decisions: Benefits eligibility, tax audit selection, sentencing recommendations, immigration adjudication, child welfare assessments, parole, employment-relevant federal background checks
- Source code and training data subject to audit by the appropriate Inspector General and by qualified independent researchers under structured access
- Public explainability requirements: Decision logic disclosed in plain language; specific reasons for adverse decisions provided to affected individuals
- Human appeal mandatory: No fully-automated adverse decision without an opportunity for human review
- Disparate-impact testing required: Pre-deployment bias audits for race, ethnicity, gender, age, disability, geography; ongoing monitoring with public reporting
Public Information Sovereignty (Data Fiduciary)
Create a public-interest data trust framework that ensures individual citizens own and control their personal data, ending the structural exploitation of citizen data by tech monopolies.
- Fiduciary duty established: Platforms holding personal data of more than 1 million U.S. users owe affirmative duties of loyalty and care to data subjects (modeled on Jack Balkin’s information fiduciary framework)
- Right to delete, port, and access: Universal — comparable to GDPR Articles 15-20 — with statutory damages for violations
- Sensitive data categories restricted: Health, location, biometric, children’s data — heightened protections; no advertising-based monetization without explicit informed consent
- No retroactive consent extraction: Data acquired without consent under prior regimes cannot be retroactively monetized
- Children’s data: Strong default protections; no behavioral profiling of minors
Platform Curation and Attention Pollution Rules
Establish structural requirements for Systemically Important Digital Platforms (SIDPs) — those reaching more than 10% of the U.S. population — to protect the public square.
- Chronological feed default: Mandate that chronological feeds (non-algorithmic, based on explicit user subscriptions) be the default state at signup, with algorithmic curation as a strictly opt-in feature that users can revoke at any time
- Algorithmic Impact Assessments: SIDPs must perform and publish pre-deployment risk assessments on an algorithm’s impact on social polarization, disinformation, and youth mental health
- Attention Pollution Framework: Treat large-scale algorithmic amplification of outrage-maximizing false content as a negative economic externality (pollution); impose a graduated excise on advertising revenue derived from algorithmically-amplified content that has been identified as inauthentic or harmful at scale
- Recommender system inspection rights: Qualified independent researchers (Stanford-modeled vetting) receive structured access to recommender system outputs for studying social effects
Structured Digital Deliberation
Integrate structured digital town hall platforms into federal public comment and rulemaking processes — using technology to surface consensus rather than amplify divisions.
- Federal Deliberative Platform: Modeled on Taiwan’s vTaiwan / Polis system, which has been used for genuine bottom-up policy formation since 2014
- Integration with rulemaking: Replaces existing “notice and comment” docket systems for selected complex rulemaking, surfacing areas of public consensus and disagreement
- Coordinated with Federal Deliberative Council (Freedom to Vote Act): online platform feeds into Citizens’ Assembly deliberation; assembly recommendations route back to platform for refinement
- No algorithmic ranking by emotion or virality: Comments surfaced by structural disagreement reduction, not engagement maximization
Algorithmic Liability for State Actors
Amends Section 230 so platforms lose liability shields if their algorithms amplify coordinated, state-sponsored information operations.
- State-sponsored disinformation carve-out: Section 230 immunity does not apply to algorithmic amplification of content originating from designated state-affiliated operations (Russia IRA-pattern, PRC United Front, etc.)
- “Verified Human Citizens” filter toggle: Mandate a user-controllable toggle on SIDPs that filters feeds strictly to verified-human accounts, throttling bot networks
- Digital Keys (optional): Cryptographic credentials proving humanity and citizenship without revealing identity — users may use them to gate access to their own content if desired
- Foreign-funded proxy provisions cross-referenced with the Total Ban on Foreign-Funded Proxies in the Government Transparency Act
Constitutional Authority
Article I, Section 8 (Commerce Clause — interstate communications and digital commerce, well-established basis for FCC, FTC, and CFPB jurisdiction); Article I, Section 8 (Spending Clause — federal IT modernization and AI investment); 14th Amendment Section 5 (Congressional enforcement of equal protection, supporting algorithmic anti-discrimination provisions); 5 U.S.C. § 552 (FOIA framework supporting algorithmic transparency); 5 U.S.C. § 552a (Privacy Act framework supporting data fiduciary rules). Section 230 carve-outs: Congress retains plenary authority to define the boundaries of Section 230 immunity, consistent with Force v. Facebook dissents and the First Amendment limits identified in Knight First Amendment Institute v. Trump (2nd Cir. 2019). Open-source software mandates: well-precedented through existing federal procurement (DoD open-source mandate, OMB M-16-21). The information fiduciary framework is consistent with existing fiduciary doctrine and Riley v. California (2014) on the constitutional sensitivity of digital data.
Rationale
Digital infrastructure has become as foundational to modern life as roads, electricity, and telephones once were — but unlike those earlier utilities, the digital layer was built almost entirely by private firms whose business models depend on surveillance, attention capture, and algorithmic optimization for engagement. The result is a deliberative substrate that systematically degrades public reasoning, an opaque algorithmic apparatus that makes consequential decisions about Americans’ lives, and a small number of private firms that wield more power over public discourse than any 19th-century newspaper baron. This Act does not propose to dismantle digital infrastructure. It proposes to govern it the way every prior critical infrastructure has been governed: with transparency, accountability, public-interest constraints on the most consequential uses, and citizen sovereignty over the data that describes their lives. AI specifically is at a moment that demands proactive governance — the trajectory of the technology is steep, the stakes for democratic governance and economic mobility are very high, and the existing regulatory state lacks both the authority and the capacity to keep pace. This Act builds that authority and that capacity before the trajectory becomes irreversible. It is not “Left vs. Right.” It is “Working vs. Broken” — and right now the digital governance layer is broken.
Implementation Timeline
- Year 1, Q1: National AI Council established by executive order; AI Safety Institute statutory authority enacted; algorithmic transparency rule effective for federal AI systems
- Year 1, Q2: Section 230 amendment effective (state-sponsored disinformation carve-out); SIDP designations begin; algorithmic impact assessment requirements effective
- Year 1, Q3: Data fiduciary regime effective; right-to-delete and right-to-port live for U.S. users; Digital Sovereign ID pilot launched (opt-in, replaces existing Login.gov)
- Year 1, Q4: Federal Deliberative Platform stood up; first rulemaking conducted on the platform; chronological feed default requirement effective
- Year 2: Attention Pollution Framework excise effective; cross-agency Digital Front Door rollout begins
- Year 3-5: Government IT modernization completes for top 50 agencies; AI workforce transition fully coordinated with Federal Job Guarantee; recommender system researcher access program institutionalized
Fiscal Impact
Total federal cost at full implementation: $18-26B annually steady-state plus $40-60B over 5 years in one-time IT modernization investment.
- National AI Council and AI Safety Institute: $2-3B annually
- Government IT modernization: $40-60B over 5 years (one-time); maintenance $4-6B annually thereafter
- Digital Sovereign ID infrastructure: $1B annually
- Algorithmic auditing capacity (across DOJ, EEOC, CFPB, FTC, ED): $1-2B annually
- Federal Deliberative Platform operations: $200M annually
- FTC Data Fiduciary enforcement: $500M annually
- Public compute infrastructure for research: $4-6B annually
Funded by: (a) Attention Pollution excise on advertising revenue from harmful algorithmic amplification ($3-8B annually estimated); (b) data-fiduciary violation penalties and disgorgement; (c) general appropriations balance. Economic returns: government productivity gains from automation ($30-60B annually at full implementation per OMB modeling), reduced fraud through better data integration ($10-25B annually), and substantial consumer welfare gains from data sovereignty rules. Net long-term fiscal benefit strongly positive.
Political Considerations
This Act faces concentrated opposition from: (a) major digital platforms (revenue model implications of chronological feed defaults, fiduciary duty, and Attention Pollution excise); (b) federal contractors heavily invested in proprietary legacy IT systems; (c) AI labs concerned about transparency and audit requirements. Polling: 70-78% support for algorithmic transparency in government decisions; 60-68% for chronological feed defaults; 65-72% for data ownership / right to delete (Pew, Wired/YouGov, 2022-2025); 75-82% for prohibiting children’s data monetization. Strongest political vulnerabilities: (a) AI governance characterized as “stifling innovation” — counter with the empirical record that consumer-protection regimes (FDA, FAA) accelerate sustainable innovation by establishing trust; (b) data fiduciary characterized as “European-style regulation hurting American companies” — counter with the polling showing strong U.S. demand for data sovereignty; (c) Section 230 carve-out characterized as a content-moderation power grab — counter with the narrow drafting limited to state-sponsored coordinated inauthentic behavior, not content viewpoint. The Digital Governance Act is foundational: without it, every other Act depends on a digital substrate that systematically undermines the conditions for informed citizen participation.