Hackers and scammers are sick of AI spam: junk has flooded their closed forums

Cybercriminals have run into an unexpected problem: a wave of AI-generated content has swamped even closed forums and the dark web. Hackers and scammers are now complaining about the glut of neural networks in the places where they try to discuss cyberattacks and illegal activity.

When spam reaches the underground

AI content is now everywhere. Neural networks generate articles, comments, posts and forum messages so fast and so cheaply that traditional platforms such as Reddit, Twitter and YouTube have long been oversaturated with the material. Now the wave has reached specialized closed forums, where cybercriminals gather to discuss their activities.

Closed cybercriminal communities were long considered a relatively safe haven. They have fewer government intelligence agencies, better moderation (among their own), and a higher average quality of information. Members of such forums discussed real hacking methods, sold stolen data and shared best practices, knowing their content would not reach outsiders. But AI spam has made its way in there too.

How low prices created the problem

• The cost of generating text through an API has fallen through the floor
• Spambots can be programmed in hours using off-the-shelf models
• Low barriers to entry for bad actors who want to clog a forum
• No specialized protection against AI automation on closed platforms
• Rapid growth in the availability of powerful language models through APIs

The problem is that neural network models are now cheap enough and simple enough to use. A spambot can be assembled in a few hours using the OpenAI API, the Claude API or freely available models on Hugging Face. Generating 10,000 forum messages now costs pennies. Spambots start flooding forums with generated content, ranging from meaningless posts that pad an activity counter to more cunning schemes in which an AI bot imitates an experienced member to earn reputation and then sell the hacked account. At the moment this is even cheaper and more effective than hiring people to do the same job.

The paradox of technology

There is a certain irony in this situation. The cybercriminal community, which itself actively uses technology and automation for its own ends, is now suffering from the consequences of the development of entirely different technologies. Hackers were among the first adopters of automation, bots, scripts and AI solutions. But nobody expected this wave to turn against them.

Historically, cybercriminal forums were several steps ahead of the mainstream internet in technical solutions. They adopted encryption, anonymity and decentralization before those concepts became popular. But AI spam is a problem that scales much faster than any traditional moderation can handle.

"Even in the underground internet there is no peace anymore," is the

gist of complaints from active members of closed forums.

What this means

The spread of AI content has already reached a scale at which it affects even those niches that try hardest to hide from it. It is a telling signal that information quality and content control are no longer a problem for big tech platforms alone, but a global reality that touches literally everyone.

Cybercriminals are being forced to look for new ways to defend against AI automation, just as ordinary users and organizations do. The irony is that they may use the same tools as their enemies: more advanced AI detection models, new user verification methods, decentralized architectures. But this is a problem they clearly did not expect to have to solve.