AI safety is still an issue with bad AI. Not only do you have to protect against what it can do, but also protect against what it might accidentally do, the lies it tells that the masses believe, or people seeding the internet with misinformation specifically for AI to pick it up and regurgitate it to people without a proper source. It’s like money laundering, but for misinformation.

All these things are safety concerns of AI, but just rogue AGI.