--- title: "Digital Rights & AI | A Government That Works for the People" canonical: "https://govthatworks.org/digital-rights" description: "Chapters 12, 32, and 33: data minimization, data brokers, government purchases, automated decision rights, synthetic media, provenance, and anti-SLAPP." --- Digital liberty · Chapters 12, 32 & 33 # A Digital Bill of Rights The digital framework tries to constrain both private data power and government surveillance while keeping political speech outside a government truth-adjudication system. This page distinguishes the book’s proposed federal statute from its proposed constitutional privacy amendment. Neither is current law. **Read the source chapter in the book**Chapters 12, 32 & 33 — Privacy, Data, AI, and Speech [Open Chapter 12](/reader?ch=12#chapter-12)[Open Chapter 32](/reader?ch=32#chapter-32)[Open Chapter 33](/reader?ch=33#chapter-33) Chapter 32 ## An American Data Rights Act built around minimization. The book’s starting rule is purpose limitation: a service may collect and use data reasonably necessary to provide what the person requested, plus narrow operational purposes defined by law. Other uses require separate, genuine consent rather than a take-it-or-leave-it privacy policy. Minimize ### Collect less by default. Access, correction, deletion, portability, and opt-out rights sit on top of a default rule that data should not be accumulated merely because storage is cheap or future monetization is possible. Sensitive data ### No ordinary market in intimate dossiers. Precise location, biometrics, health information, sexual or intimate information, religion, immigration status, and children’s data receive heightened restrictions. Data brokers register, disclose sources and customers, and provide centralized deletion. Government ### Do not buy around a warrant. The proposed statute bars federal agencies from buying commercial personal data that would otherwise require legal process to compel, complementing the book’s proposed constitutional warrant rule for sensitive information. Automated decision rights ## A consequential algorithm must be answerable to a person. For decisions involving employment, credit, housing, education, insurance, health care, public benefits, or liberty, the book requires plain-language notice that automation was used, meaningful disclosure of the principal factors, a way to correct input data, and timely human review. ### Notice Tell the affected person when a consequential decision substantially relied on an automated system. ### Explanation Identify the principal factors that drove the outcome in language a person can use. ### Correction Allow inaccurate or incomplete input data to be challenged and corrected. ### Human review Provide a timely appeal to a human decision-maker with authority to change the result. Chapter 33 ## Free speech without a federal “truth ministry.” The book rejects government adjudication of political truth. Its election-synthetic-media approach is disclosure-based: if a materially deceptive synthetic communication depicts a candidate saying or doing something the candidate did not say or do, the communication carries clear notice. Provenance standards can supplement disclosure by carrying tamper-evident creation and edit history. ### Synthetic-media disclosure Require conspicuous disclosure for deceptive synthetic election communications rather than banning controversial political claims. The page identifies C2PA Content Credentials as one possible technical implementation for cryptographically verifiable provenance, not as proof that content is true. [C2PA specification](https://spec.c2pa.org/) ### Federal anti-SLAPP protection Create an early dismissal procedure and fee shifting for suits targeting speech on matters of public concern, with safeguards so genuine defamation, fraud, threats, and other unlawful conduct can still proceed. **Goal** Give people enforceable control over personal data and meaningful recourse when automated systems make consequential decisions, without appointing government as an arbiter of political truth. **Instrument** American Data Rights Act; high-impact AI duties; data-broker registration and deletion; restrictions on sensitive-data sale; government purchase limits; election synthetic-media disclosure and provenance; platform process transparency; federal anti-SLAPP. **Who implements** FTC and state attorneys general for consumer data; NIST and sector regulators for high-impact AI; federal courts for anti-SLAPP; election regulators for campaign disclosure. **Legal vehicle** Federal statute for the Chapter 32 and 33 framework; proposed Amendments One and Five would constitutionalize selected election, privacy, and automated-decision protections. **Principal risk** Regulatory “risk” and transparency duties become indirect pressure on platforms or speakers to suppress lawful political speech. **Safeguard** Duties are framed around process, disclosure, privacy, fraud, and appeal rights; the book calls for statutory restrictions on regulator coercion of lawful speech and ordinary judicial review. The Strongest Counterargument & Our Response ### Best case against The strongest civil-liberties objection is that a broad digital regulator can become a speech regulator in practice. “Risk assessment,” provenance, researcher access, and synthetic-media rules may create levers officials can use to pressure platforms to demote lawful viewpoints, while data-minimization rules can also reduce useful innovation. ### The author's response The book draws a boundary between regulating conduct and deciding truth. It favors data minimization, fraud and impersonation rules, appeal rights, disclosure, and auditable procedures while rejecting a government power to label ordinary political claims true or false. Provenance is treated as information about source and editing history, not a truth certificate. ### What remains uncertain The line between process regulation and indirect speech pressure is fact-dependent. The effectiveness of provenance when credentials can be omitted, stripped, or ignored, and the cost of compliance for smaller firms, would need continuing evaluation. **Read the source chapter in the book**Chapters 12, 32 & 33 — Privacy, Data, AI, and Speech [Open Chapter 12](/reader?ch=12#chapter-12)[Open Chapter 32](/reader?ch=32#chapter-32)[Open Chapter 33](/reader?ch=33#chapter-33)