0:00
In Hessen we take the weakened variant, in the East you can rather hit it. This is the statement of an AfD politician when he speaks about the internal AI tool of the AfD. But what happened? The research network CORRECTIV uncovered undercover that the AfD works with an independently built AI tool.
0:19
and that's distributed in their circle of friends all over Germany. The tool is called "Alltanita Studio" and belongs to Hans Holger Maikomess. And it's not just some dude, but he's the head of the federal central office of the AfD. That means the tool belongs to a company that the AfD itself belongs to. That's interesting to know, but it gets much more interesting if you look at how the tool works. Every 15 minutes, this tool spits out messages from specially selected websites.
0:47
Then the AI tool evaluates every message with a score. And not how relevant the topic is or how much truth is in it, but how much anger and outrage can be made of it. But the whole thing goes on. Because from one such message, this tool then builds ten different AI-generated, finished social media posts. And for all possible platforms. And now comes the quote from the beginning to be carried. Because regionally it can also be adjusted to how far you
1:15
these things. So for Hesse you want the more degraded variant and in the east you can then easily hit each other with very violent blows. And of course you can now go around and say: the other parties are just to blame for not having such a content machinery in the background. You always have to be on social media as much as possible
1:34
post too much to stay relevant. And no, I don't think so. This is about politics and that's an absolute "empowerment factory" you can almost call it. This whole tool is just designed for it and only works if you don't care what comes out of it in the end. So here it's just about triggering any negative emotion. And it would be absolutely
1:54
It would be absolutely catastrophic if every party now jumps on this train and does something similar, because where do we end up? And maybe it's also very interesting to know that this whole tour works with interfaces from Google, Anthropic and OpenAI. And yes, it is actually the case that Chachibiti, Gemini and Claude do not even allow such political campaigns.
2:16
If you pay directly for the interface, you are basically bypassing the chat window and thus the guidelines and simply only use the capabilities of the language model. You have to know for yourself whether you want to vote for a party that simply builds on anger.
