OpenAI Supports California Senate Bill 1119 for Youth AI Safety

OpenAI supports California Senate Bill 1119 (SB 1119), a legislative effort to establish mandatory safety safeguards for young people using AI. This support is intended to create a strong standard for youth AI safety in the absence of federal action, ensuring that teens have access to AI tools for learning and productivity while maintaining strong default protections.

Key Requirements of SB 1119

Senate Bill 1119 proposes a framework that combines accountability with age-appropriate protections. OpenAI specifically supports the following requirements outlined in the bill:

  • Age Verification: The ability to determine a user's age.
  • Risk Assessment: Identifying and addressing safety risks before making a product available to young people.
  • Independent Audits: Requiring products to undergo independent audits.
  • Harmful Content Protection: Protecting young people from high-risk interactions and harmful content, including sexually exploitative content and self-harm.
  • Parental Oversight: Providing parents with tools to guide and limit their children's AI usage.
  • Crisis Support: Connecting young people with crisis-support resources when serious safety risks are detected.
  • Data Privacy: Limiting targeted advertising and protecting the personal information of young people.

For users identified as being between the ages of 13 and 17, these protections are designed to apply automatically.

ChatGPT for Teens and Built-in Safeguards

OpenAI has implemented a "ChatGPT for Teens" experience designed for users under 18. If a user states their age is between 13 and 17, or if the system estimates they are under 18, they are automatically placed into this specialized experience. These protections are baseline requirements rather than optional settings.

Educational Tools

ChatGPT for Teens includes features developed with educators and learning experts to support critical thinking and study habits, such as:

  • Quizzes
  • Learning Visualizations
  • Study mode (designed to guide students through problems step-by-step)

Safety and Contextual Continuity

OpenAI utilizes context and memory features to apply safeguards consistently and recognize concerning patterns over time. This continuity allows teens to engage in long-term learning projects without restarting and helps the system provide more effective responses when a young person requires additional resources.

Ongoing Safety Frameworks and Principles

OpenAI has integrated several safety measures into its product development and Model Spec, including:

  • Parental Controls: Tools for families to manage settings and receive notifications in high-risk situations.
  • Age Prediction: Systems to estimate user age to apply appropriate protections.
  • Teen Safety Blueprint: A strategic framework for youth safety.
  • Model Spec Guidelines: Rules that prohibit the AI from engaging in romantic relationships, encouraging emotional dependence, or claiming to be human or sentient.

OpenAI views SB 1119 as a framework that can be improved through continued study and collaboration during its implementation to keep pace with evolving AI technology.

Sources