California State Senator Wiener proposed SB 1047 to regulate the risks associated with AI. The bill, if enacted, would require AI developers to determine if an AI model has a “positive safety determination," which would be defined to mean that the developer has a reasonable belief certain types of AI models to not have a hazardous capability, or come close to having a hazardous capability when accounting for margin of safety and post-training modifications.
- Security protocol (including a separate safety and security protocol meeting certain specifications)
- Implementing reasonable safeguards to prevent users from using the hazardous capabilities of the AI model.
- Refraining from making use of the model if there is an unacceptable risk of use of the hazardous capabilities.
While the law has a long way to go before enacted into law, model developers may wish to consider the policies and procedures proposed by the bill to minimize risk associated with the use of AI.