An algorithm denied your loan, screened you out of the job, set your bail, flagged your face. You were never told, never shown the reasoning, and had no one to appeal to — because "the system decided." This statute ends the black box: you have the right to know a machine decided, to challenge it, and to have a human being answer for it.
An algorithm scored you — for a loan, a job, an apartment, an insurance rate, a bail hearing — and no one had to tell you it happened, let alone why.Check it
The models were trained on a history of discrimination and reproduced it faithfully, at scale, while everyone involved pointed at the machine and shrugged. No human was accountable.Check it
Synthetic audio and video of real people flooded elections with no disclosure, and facial recognition drove arrests of people who had done nothing.Check it
The most powerful systems ever built were developed with no safety threshold, no oversight body, and no one empowered to say stop.Check it
You have the right to know when an AI system made or shaped a decision about you, to see the basis for it, to challenge it, and to get human review.
High-stakes AI — employment, housing, credit, insurance, government — faces mandatory bias audits. Discrimination laundered through a model is still discrimination, and the developer and deployer are liable.
In criminal justice, AI faces hard prohibitions — a machine doesn’t get to decide your liberty in the dark.
AI-generated audio and video must be disclosed, election deepfakes are addressed, autonomous weapons need Congress, and an independent commission sets binding safety thresholds on frontier systems.
No independent body had authority over the AI systems making consequential decisions about people’s lives.Check it
An independent AI Safety and Governance Commission — seven commissioners, seven-year staggered terms, no more than four from one party, removable only for cause. Its authority reaches federal AI, AI affecting individual rights, and AI in critical infrastructure.
Models trained on discriminatory history reproduced it at scale, and "the algorithm did it" worked as a defense.Check it
Mandatory bias audits for government and high-stakes private AI. Bias laundered through a model is still bias — and it must be found, disclosed, and fixed.
An automated score could sink you with no notice, no explanation, and no appeal to a human being.Check it
Any entity using AI in employment, housing, credit, or insurance decisions owes you notice, the basis for the decision, and a route to challenge it with human review. And you get a private right of action to enforce it.
Risk scores and facial recognition shaped bail, charging, and arrests — opaque, unchallengeable, and wrong often enough to ruin lives.Check it
Specific prohibitions and requirements govern AI anywhere in a criminal process. A machine’s output cannot quietly become the reason you lost your freedom.
Synthetic audio and video of real people spread with no label, and arrived in elections with nothing to stop them.Check it
AI-generated audio, video, and images require disclosure, with specific election-integrity requirements. You get to know whether the thing you’re watching is real.
Liability evaporated between the developer, the deployer, and the machine — leaving the injured person with no one to sue.Check it
A developer or deployer is liable for the AI’s content and decisions. Accountability lands on a person or a company — never nowhere.
The most capable systems ever built raced ahead with no capability thresholds — and lethal autonomy crept in without a vote.Check it
The Commission sets capability thresholds and frontier-AI safety requirements. Autonomous weapons require congressional authorization — no machine takes a human life on an executive’s say-so.
This is the plain-language version. The binding text is CS-20, which implements Article Twenty-Two of the Constitution.
This is your country’s law. Help shape the next draft of it.
