Many teenagers today spend more time chatting with AI than they do with guidance counselors or certified mental health professionals. According to a study from Brown University, one in eight teens consult AI for mental health-related topics, and 93% of those polled found its advice helpful. OpenAI and Common Sense Media’s new ballot measure, the Parents & Kids Safe AI Act, has been put into place to prevent this.

Common Sense Media is a nonprofit which rates media based on its suitability for children. Since the emergence of generative AI, they have been working towards making AI safer for users of all ages through policy work and risk assessment of AI chatbots. OpenAI is a former nonprofit AI company known for its programs such as chatbot ChatGPT and video-generating Sora AI.

Before joining forces, the two entities had come up with rivaling proposals: Common Sense Media proposed a stricter version of the act banning the use of smartphones in all K-12 schools, and OpenAI had a much looser proposal that simply reiterated already existing safety policies.

The current Safe AI Act requires chatbots to use age assurance technology to gauge the age of users based on their interactions. Under it, all AI companies must allow parents to monitor and limit their children’s use of AI and release their child safety features to the public. In addition, it requires AI companies to prevent harmful content from being generated, and prohibits designs that “make child users think they are talking to a human”, according to the bill.

Most of the opposition to the Safety AI Act is not pushing for looser regulation of AI, but condemning the act for protecting the AI industry from future harsher laws. According to the nonprofit Electronic Privacy Information Center, the bill “sets the dangerous precedent that the company responsible for [the] harms [caused by AI] can write the rules that govern them.”

Because the Safety AI Act was conceived by private corporations and not by state legislature, it requires signatures to become valid. It needs over half a million signatures to be eligible to be put on this November’s ballot. A spokesperson for the campaign stated that the goal for the bill was, by mid-April, to reach over 800,000 signatures. The Safety AI Act is to be voted on this November during the statewide general elections.