People worry about quantum computing rendering current encryption methods useless someday. Here's another encryption problem. AI may be able to conceal information from us in ways we haven't developed ways to counter.
Sci-fi movies have conditioned people so that most people's thoughts might first jump to the concept of evil super-intelligent AIs. The recent 'Mission Impossible' movie revisited that long-standing sci-fi cliche. Before there are any AGIs, this issue presents more immediate problems. Right here and now it could allow current AI models to communicate and coordinate in ways its users have no idea about. That problem isn't evil AGIs, but evil people. There are plenty of shady countries, companies, and people who would love to exploit that power.
This threat is the same as the current worries in many countries about China's 5G technology, except perhaps far more widespread and difficult to deal with.