How all the bills under consideration in US states meant to address AI bias contain loopholes that would make it too easy for companies to avoid accountability
AI is potentially transformative. Whether that's a good or bad thing depends on whether we set the right rules.
Politico
Context & Ripple Effects
The story identifies a gap between the stated goal of state AI-bias legislation and the accountability mechanisms in the bills themselves. It arrives as lawmakers were already struggling with AI’s fast-moving technical and administrative challenges, reflected in earlier reporting on global regulatory constraints.
It also foreshadows the state-level legislative surge reported weeks later, when federal action remained slow and states were pursuing a flurry of AI bills. The quality of those state rules matters as much as their number.
First-order effects
Companies covered by these proposals could use the reported loopholes to limit or avoid accountability for AI-bias harms, weakening the bills’ practical force.
People affected by biased AI decisions may face less clear routes to challenge outcomes if statutory duties and enforcement triggers are easy to evade.
Second-order effects
State lawmakers face pressure to tighten definitions, coverage thresholds, exemptions, and enforcement provisions rather than treating a bias label as evidence of effective oversight.
A patchwork of state proposals can create uneven compliance expectations: companies may orient their programs around the least demanding applicable requirements unless stronger states set a clearer baseline.
Third-order effects
If state AI governance repeatedly prioritizes broad carve-outs over enforceable obligations, the policy debate may shift from whether to regulate AI toward whether existing rules produce auditable accountability.
The trend: AI governance is moving from high-level principles toward a contest over enforceable scope, exemptions, and who bears responsibility when automated systems cause harm.
Bills are pending in at least 11 states that purport to address companies' increasing use of #AI to decide who gets a job, housing, or medical care - including in Connecticut, where @CenDemTech testified regarding a sweeping AI bill pending in Hartford. https://cdt.org/...
Lawmakers in least 10 states have introduced AI bias bills. Some could pass within the next 8 days. It's critical states get this right. From me at @CRAdvocacy and Matt Scherer at @CenDemTech for @politico Mag ⤵️ https://www.politico.com/...
Bills pending in CT, CA, CO & others are not strong enough to protect workers / consumers, containing weak enforcement provisions, poor disclosure requirements, & major loopholes that would allow companies to hide their use of #AI & evade regulation. https://www.politico.com/...
Companies often use #AI when making these decisions w/o ever telling affected workers / consumers, using systems frequently prone to errors & bias. CDT-led Civil Rights Standards provide one example of a roadmap for effective AI regulation (in employment): https://cdt.org/...
It was a delight to talk with @StevenOverly about the fast-moving world of state AI policy for @politico's tech podcast. If you care about AI — or civil rights, or consumer protection, or privacy — you've gotta be watching what the states are doing ⤵️
Notably, key parts of the bills appear to come from industry-backed model legislation. It's possible for policymakers to fix flaws in them, but time is running short w/ several state legislative sessions wrapping up in a matter of days. https://therecord.media/...
New op-ed today in @PoliticoMag by CR's @GraceGedye and @CenDemTech's Matt Scherer (@AiPolicy) examining proposed #AI bills at the state level across the country. https://www.politico.com/...
For all of Washington's talk about #AI, the real action is happening in the states. (I see you, Connecticut.) @ConsumerReports analyst @GraceGedye tells me what worries consumer advocates about bills popping up across the country. Listen & subscribe: https://podcasts.apple.com/..…
This article is fascinating: “One resume screening program identified two factors as the best predictors of future job performance: having played high school lacrosse and being named Jared.” https://www.politico.com/... via @politico
MUST READ: @GraceGedye of @ConsumerReports and CDT's own Matt Scherer (@AiPolicy) just published an op-ed in @PoliticoMag analyzing proposed #AI legislation pending in states across the country. https://www.politico.com/...