Civil Rights Letter to House Democratic AI Commission

The Leadership Conference on Civil and Human Rights, representing over 200 organisations, has written to the House Democratic AI Commission demanding mandatory bias testing, privacy protections, and disparate impact liability. This is a perfectly reasonable set of demands directed at an advisory body whose main power is to produce a report that will be read by approximately twelve congressional staffers. The letter notes that existing AI legislation is 'incomplete without provisions that hold developers accountable for discriminatory outcomes'—a polite way of saying most current bills are glorified suggestion boxes.

The three demands are the civil rights equivalent of a triple espresso: mandatory bias testing (which tech companies will gamify into submission), privacy protections (a concept Congress has been 'working on' since the Clinton administration), and disparate impact liability (the one with actual teeth, and therefore the one most likely to be lobbied into oblivion). The letter even references Title VII of the Civil Rights Act as a model, apparently forgetting that it took fifty years of litigation to make Title VII stick against human discrimination. Expecting it to restrain algorithms with loss functions is optimistic.

The House Democratic AI Commission was created as a pressure-relief valve, not a reform engine. It collects letters like this, files them, and eventually issues a report with recommendations that Congress will politely ignore. The Leadership Conference knows this; the letter is partly a historical document, a paper trail for the inevitable lawsuit when someone's AI denies housing or employment based on patterns it has learned from biased data. 'We tried,' the letter will say in the footnotes of litigation. And they did. But trying and succeeding are two different verbs, and Washington only seems interested in the first.