DeepTrust AI Transparency Scoring Law Takes Effect in California: High-Risk AI Systems Must Disclose Credibility Metrics

DeepTrust AI Transparency Scoring Law Takes Effect in California: High-Risk AI Systems Must Disclose Credibility Metrics

California Senate Bill SB-1127 has come into force. The law requires every high-risk AI system operating in California to disclose a credibility score to users, with violations punishable by fines of up to $75,000 per case. It is the first US state law to impose a mandatory disclosure obligation on AI credibility.

The scope of "high-risk AI systems" is defined by the California AI Safety Board (CalAISB) and covers AI used in recruiting, medical diagnosis, credit decisions, judicial assistance, and autonomous driving. These systems must compute a credibility score under CalAISB's unified standard and present it graphically in their user interfaces.

The credibility score has eight dimensions: accuracy, explainability, auditability, bias level, robustness, privacy protection, energy efficiency, and emergency control capability. Each dimension is scored from 0 to 100, and the system produces a composite credibility score. CalAISB publishes a detailed evaluation toolkit; developers can self-score or use a third-party assessor.

Enforcement is handled by the Technology Enforcement Division of the California Department of Justice. The first enforcement actions will target three categories: resume-screening systems at online hiring platforms, small-business credit approval systems at banks, and clinical decision support systems at medical institutions. Operators have 90 days from the law's effective date to complete their first compliance assessment.

Supporters argue the rule will lay a foundation for trust in AI markets. Deirdre Mulligan, an AI ethics professor at the University of California, Berkeley, said users have a right to know the credibility of AI decisions that affect their lives. Mandatory disclosure, she argued, will push AI providers to actively improve system quality, creating a virtuous competitive dynamic.

Opposition has come mainly from the AI industry. Tina Park, executive director of the Silicon Valley AI Alliance, said the law is unfair to small and mid-sized AI companies — large firms have the resources to run sophisticated credibility assessments, while smaller players may exit the California market because of compliance costs. The view has been echoed by some members of the California Chamber of Commerce, which has jointly filed an amendment proposing that the compliance threshold be raised to companies with more than 100 employees.

Internationally, SB-1127 aligns closely with parts of the EU AI Act. Margrethe Vestager, the European Commission's executive vice president for digital affairs, said the law's entry into force proves AI regulation has become a global trend and that she looks forward to sharing best practices with California regulators. Japan, South Korea, and Brazil are also considering similar legislation.

The first compliant AI systems are already on the market. Resume-screening platform HireVue has completed its credibility assessment with a composite score of 84. HireVue CEO Kevin Parker said proactively disclosing the score lifted user trust in the platform by 23 percent and registration conversion by 15 percent.

The objectivity of the scoring methodology remains contested. CalAISB's framework leans heavily on technical metrics and does not pay enough attention to social impact, according to critics. Lisa Macpherson, policy director at the public-interest group Public Knowledge, suggested the system should add dimensions for environmental impact assessment and social benefit assessment.