Earlier this year, a US entrepreneur set up a social network for AIs. The network is called Moltbook and the types of machines on there are called AI agents, which can perform tasks that go beyond chatbots like ChatGPT. The AIs were given personalities by humans but were then left to interact with each other. Humans were allowed to observe but not intervene – at least in theory. That’s when things really got weird.

AI researcher David Reid at Liverpool Hope University highlights instances of AIs creating their own religions, operating a marketplace for digital “drugs” and evading human efforts to eavesdrop on their conversations. This comes with a health warning: there is evidence that humans have infiltrated the site and are posing as AI agents. However, if some of the activity on Moltbook can be attributed to machines, it may hint at the development of complex “emergent” behaviour that doesn’t come from their programming.

A recent decision by officials means the NHS will soon be allowed to spend much more on new drugs. But who will it really benefit? And should you be practicing the ventilation technique of “house burping” – or Stoßlüftenas it’s known in Germany where it’s long been popular? An expert in indoor air pollution looks at the evidence.