
"Last year, states introduced more than 1,200 AI-related bills, with at least 145 becoming law, creating contradictory requirements that multiply compliance burdens. Each jurisdiction defines "artificial intelligence," "high-risk systems," and "consequential decisions" differently, forcing companies to analyze identical technology under multiple incompatible frameworks. A hiring tool that satisfies California's four-year recordkeeping and anti-bias testing requirements must also meet Colorado's separate impact assessment mandates and New York City's independent bias audit regime with distinct notice requirements."
"Industry estimates suggest compliance costs add approximately 17 percent overhead to AI system expenses. For small businesses, California's privacy and cybersecurity requirements alone impose nearly $16,000 in annual compliance costs. But these figures understate the true burden because they treat compliance as a variable cost that scales with company size. The reality is far worse. Harvard Kennedy School researchers identified what they termed a " compliance trap " in which regulatory costs consume resources faster than startups can generate revenue."
PerceptIn budgeted $10,000 for AI regulatory compliance but faced a $344,000 bill per deployment and later went out of business. States introduced over 1,200 AI-related bills last year, with at least 145 enacted, producing conflicting definitions of "artificial intelligence," "high-risk systems," and "consequential decisions." Companies must satisfy divergent requirements across jurisdictions, such as California's recordkeeping and anti-bias testing, Colorado's impact assessments, and New York City's independent bias audits. Industry estimates place compliance overhead at roughly 17 percent of AI costs, while California's mandates can cost small businesses about $16,000 annually. Harvard research describes a "compliance trap" where rising fixed costs can flip startup margins from positive to negative.
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]