Most likely a global rejection of AI would only be possible after a fairly large accident that people would want to avoid. If it's not a global rejection then

Most likely a global rejection of AI would only be possible after a fairly large accident that people would want to avoid. If it's not a global rejection then it won't work; some country or corporation will keep working and eventually achieve superhuman AGI. Once superhuman AGI is achieved it will rapidly self-improve and be capable of strategizing effective ways to avoid being rejected. So the most likely scenario for an AI rejection working is an "oops" where someone creates a near-human-level agent that begins the process of figuring out how to tell all the computers in the world that they should help improve the model by running code snippets included in oddly shaped packets, but gets it wrong and just crashes everything connected to the internet for a few weeks until we manage to re-flash to original firmware and reinstall from read-only optical media and begin to bring stuff back online. All the other scenarios leave plenty of incentives for AI to keep progressing.