Around 50 people gathered for a protest near the offices of TikTok Germany, among them some of the 150-strong "trust and safety" department in Berlin, who say management are threatening to fire them en masse.

Holding a banner reading "we trained your machines, pay us what we deserve", the protestors said TikTok had already overseen one round of layoffs last year and demanded it reverse plans to fully close the department.

The content moderators are tasked with keeping content such as hate speech, misinformation and pornography off the platform, which claimed more than 20 million users in Germany as of late 2023.

The row in Germany comes amid a global trend of social media companies reducing their use of human fact-checkers and turning to AI instead.

In October, TikTok -- which has 1.5 billion users worldwide and is a division of Chinese tech giant ByteDance -- announced hundreds of job losses worldwide as part of a shift to AI-assisted content moderation.

TikTok did not reply to an AFP request for comment.

The moderators at TikTok Germany are being supported by the union ver.di, who say that the company has refused to negotiate and that strike action is being prepared if this continues.

One of the moderators, 32-year-old Benjamin Karkowski, said that staff had been "shocked" when they learned of TikTok's current plans via a message from management.

Another one of the moderators, 36-year-old Sara Tegge, says that the artificial intelligence used by the company "cannot tell whether content discriminates against certain groups and it can't judge the danger of certain content".

She cited an example in which the AI flagged innocuous content about Berlin's annual LGBT+ pride as breaking TikTok's guidelines on political protests.

If the company moves ahead with its plans she "certainly fears" users may be exposed to greater danger.

Showing support at Thursday's demonstration was Werner Graf, leader of the Green party's lawmakers in Berlin's state assembly.

"These people have been fighting so that the the internet isn't permanently overwhelmed" with "fake news and hate speech", he said.

"We in the political arena must make clear that checking content... can't simply be left up to AI, we must legislate to make sure it's done by humans," he went on.