As a generally pro-AI person, I worry that AI will make it so that common people can never resist their governments again

[Addendum so this post doesn't get removed: This isn't a fearmongering post about AI (per rule 5) - I am pro AI and this post is in favor of more democratized access to AI capabilities for the average person.] Consider the current reality: World Superpower 1 of 2 (USA): OpenAI and Claude get to decide who has access to their models' cybersecurity capabilities. They have defense agreements with the US government so CIA/NSA almost certainly has access to them. On the other hand, these companies are under no obligation to provide the same cyber capabilities to the average citizen. For example, OpenAI and Anthropic are blocking people from reverse engineering things using their models. Someone recently got banned for reverse engineering their Samsung device to expose microphone eavesdropping capabilities. TLDR: The US government gets to find exploits in your software, but you can't protect yourself. World Superpower 2 of 2 (China): The Chinese government is enacting a policy to ban open source models from having cyberattack capabilities. But make no mistake - the Chinese AI companies' proprietary models do have cyber capabilities, but you can only access them via API. Similar to the US