United States passes new law requiring AI companies to be more transparent

California signed SB 53, one of the first significant regulations in the United States requiring large artificial intelligence companies to be more transparent about how their systems work.

What does the law require?

AI companies exceeding certain thresholds of users or revenue must:

• Publish security reports: Details on how they test their models for risks before launching them.
• Report serious incidents: If an AI system causes significant harm, the company has 72 hours to report it to authorities.
• Allow auditing: Regulators can review the security systems of the largest companies.

Whistleblower protection

The law also protects employees of AI companies who want to report risks or problems. If a worker reports a security flaw internally or to authorities, they cannot be fired or penalized for doing so. It's the first law of its kind in the United States.

Who does it affect?

The law applies mainly to companies with over 100 million active users or that train models costing more than $100 million. That includes OpenAI, Google, Anthropic, Meta and other large AI companies.

The context

California has been the most active state in regulating technology in the United States. This law adds to a package of regulations that include digital data protection and algorithmic transparency. The AI industry has lobbied intensely against these regulations, arguing they stifle innovation.