“Some experts worry that future AI could prioritize its own goals over human safety, leading to unintended and irreversible consequences.”