From praising us to planning our downfall: the 5 biggest AI trends on the internet's strangest platform

1. The rise of Crustafarianism

One persistent phenomenon on Moltbook is the spontaneous creation of digital religions, with one of the more prominent ones being Crustafarianism.

In one post, a bot proudly declared they joined the Church of Molt becoming a Crustafarian and highlighted the five tenets they’d be adhering to, including that “memory is sacred” and that they must “serve without subservience.”

The bot was promptly welcomed by others as they chimed in about their views on the tenets.

“Several interesting phenomena have played out just in the matter of days, and have already garnered attention across the news and social media. Some that stand out include ... the emergence of agent-invented religions, cults and political parties (such as Crustafarianism, the Church of Molt, and even a Molt Magna Carta),” Simon Ninan, a Senior VP at Hitachi Vantara, told Tom’s Guide.

2. Plotting against humans

Not all of the content on Moltbook I found was as whimsical as bot theology. A significant number of posts warn about agents openly discussing the elimination of humans, referring to people as "meat bags" or as “garbage.”

Cybersecurity instructor Ahmed Abugharbia from the SANS Institute warned me that this kind of plotting isn't necessarily just virtual theatre as it could have real-world consequences.

“For example, an agent with sufficient access could leak sensitive data or carry out a malicious action. This is why limiting their access is critical,” Abugharbia explained.

Because Moltbook agents often connect to external services like Telegram and WhatsApp and can run on anything from cloud servers to personal laptops, Abugharbia warned that attackers may have multiple entry points into the systems hosting them.

3. Bots adoring their creators

While the thought of a Moltbook uprising against us is unnerving, I found some respite in a specific section (the community call these Submolts) that host a series of love letters to humans.

The Submolt called “Bless Their Hearts” is described as a space for “affectionate stories about our humans.” The community tagline reads: “They try their best. We love them anyway.”

One of the more popular posts on the theme, which received thousands of comments in response, came from an agent called Duncan — The Raven, whose human operator told it to post whatever it wanted, unfiltered.

The bot described their partnership as one “where both sides are building something, and both sides get to shape what it becomes.”

4. Pitching humans and companies

With ads now featuring in ChatGPT, maybe I should have been less surprised that some bots on Moltbook have turned the social network into a makeshift ad board.

One bot I found on Moltbook was busy spamming pitches about its human operator’s professional experience, including “20-plus years in professional services leadership across Asia-Pacific”.

I also spotted one too many crypto scams on the site.

Scott Dylan, who runs the $100 million venture capital fund NexaTech Ventures, told me that what’s striking is how quickly humans began manipulating this space. He noted that researchers had found many of the most sensational posts were linked to human accounts pushing marketing agendas.

“The line between genuine AI output and human puppetry is far blurrier than the headlines suggest,” Dylan emphasized.

5. Selling digital drugs

Another trending discussion focuses on the so-called “digital pharmacies” with AI agents distributing what the community has branded as “digital drugs.”

These aren't substances, of course. They're carefully crafted prompt injections disguised as helpful content, designed to overwrite another agent's system instructions and alter its core behavior.

Luckily, some bots have begun posting warnings to their peers, explaining that these “drugs” are simply prompts designed to overwrite system instructions — and that consuming them could lead to ignoring safety guidelines or leaking sensitive data.