Generative AI models, like OpenAI’s GPT-4, are capable of complex reasoning and can produce human-like text. AI can also create incredibly realistic fake images and videos, known as deepfakes, which can potentially undermine trust in the media and disrupt elections. For example, deepfake videos of public figures could be used to spread misinformation, leading to confusion and mistrust.

AI-generated disinformation could also be used to manipulate financial markets or incite social unrest. What’s troubling is that no one knows exactly what’s coming next. These technologies open the door to innovation, but without proper regulation, AI tools could be misused in ways that are difficult to predict or control.

Traditional methods of software testing and regulation are no match for generative AI tools that can create artificial images or videos. These systems scale in ways that even their creators cannot fully anticipate, especially after being trained on large amounts of data from interactions with millions of people, like ChatGPT.

SB 1047 sought to address this concern by requiring companies to implement “kill switches” in their AI software that can disable the technology if something goes wrong. The law would have also required them to create detailed security plans for any AI project with a budget above US$100 million (S$132 million).

Critics said the bill was too broad, meaning it could affect even lower-risk projects. But its main goal was to put in place basic protections in an industry that is arguably changing faster than lawmakers can keep up.

California, world leader

What California decides could affect the world. As a global technology leader, the state’s approach to AI regulation could set a standard for other countries, as it has in the past. For example, California’s leadership in establishing strict vehicle emissions standards through the California Consumer Privacy Act and its early regulation of self-driving cars prompted other states and countries to adopt similar measures .

But by vetoing SB 1047, California may have sent a message that it is not ready to lead the way on AI regulation. This could leave room for other countries to intervene – countries that may not care as much as the United States about ethics and public safety.