Introduction

California lawmakers approved Senate Bill 53 in September 2025, a comprehensive AI safety and transparency law that would require large AI companies to adopt safety policies, publish model cards, report incidents to the state, and submit to independent audits and risk assessments. The bill aims to advance AI regulation, AI safety, and AI transparency in a way that could influence national policy. Governor Gavin Newsom retains veto power, so the final outcome remains uncertain.

Background on AI oversight

As AI systems grow more capable and more widely used, policymakers are focused on how to balance innovation with public safety. California has a history of shaping technology policy and now seeks to apply that influence to AI governance. Advocates call for proactive measures to manage algorithmic bias, privacy risks, and system failure, while some in industry warn about increased compliance costs and operational impacts.

Key requirements of SB 53

Transparency and documentation

• Publish model cards that explain model capabilities, limitations and intended uses to improve AI transparency and public trust
• Maintain and disclose safety policies that describe how companies prevent misuse of AI systems
• Mandatory incident reporting to state authorities when AI systems cause or contribute to significant harm

Oversight and accountability

• Independent third party audits to assess safety practices and compliance with AI governance standards
• Regular risk assessments for models with advanced capabilities to support AI risk management
• Potential liability for companies that fail to meet the new requirements

Additional protections and research access

• Whistleblower protections for employees who report AI safety concerns
• Proposals to create a public compute resource to broaden researcher access for safety studies
• Requirements targeted at companies building or deploying large scale AI models with high compute needs

How SB 53 connects to broader regulatory trends

SB 53 aligns with global moves toward formal AI management frameworks such as the NIST AI RMF and ISO 42001. Organizations preparing for compliance should consider mapping SB 53 requirements to their internal AI governance, documentation, and risk management processes. Comparisons to the EU AI Act and other state laws show a growing emphasis on transparency, explainability, human oversight and enforceable accountability.

Implications for industry and researchers

If enacted, SB 53 could change how AI companies operate in areas from product development to legal and compliance. Key implications include:

• Increased compliance costs for safety teams, documentation and audit readiness, which may favor firms with larger compliance budgets
• Greater democratization of information about AI models through model cards and public disclosures, enabling researchers and policymakers to better assess risks
• New pathways for independent safety research via proposals for public compute resources and clearer reporting of incidents

How businesses can prepare now

Organizations should treat SB 53 as a model for future AI compliance expectations and start by strengthening AI governance now. Practical steps include:

• Adopt or refine AI transparency practices by drafting comprehensive model cards and safety policies
• Implement regular risk assessments and align them with recognized frameworks such as NIST AI RMF and ISO 42001
• Prepare for independent audits by documenting development processes, training data provenance and mitigation measures for bias and privacy risks
• Establish incident reporting protocols and internal whistleblower channels to surface concerns early

Questions readers may have

How is California's SB 53 affecting AI regulation in 2025? What are the best practices for AI safety and transparency in the US? How do model cards improve compliance and trust? These user intent questions are central to understanding the practical impact of the bill and what organizations need to do to adapt.

Conclusion

SB 53 represents a significant attempt at comprehensive AI governance at the state level, focusing on transparency, accountability and risk management. While Governor Newsom's decision will determine whether the bill becomes law, the proposal already signals how regulators may expect companies to manage AI safety and compliance going forward. For businesses, researchers and policymakers, SB 53 is a reminder to prioritize responsible AI governance and to prepare for a landscape where transparency and auditability are core requirements.