The backlash to effectively censoring social media by outing bigots and haters has pushed them underground. Now tech-savvy racists and trolls have turned to anonymous chatbots and ‘intelligent’ software systems to spew their invective into the public domain.

Ironically, they were handed the idea by Microsoft’s own experiment with a naïve AI bot on Twitter called Tay, who was designed to learn from conversations with others. Very soon, it had been trained by cynical nerds to “love Hitler” and drop the F-word with abandon.

With increasing outrage at misogynistic “trumpist” remarks, racial slurs and everything-aphobic rants on Facebook, Twitter, Instagram and Pinterest, it was no longer safe for social deviants and professional offenders to ply their trade in the open. Even John Cleese found that political incorrectness might be humorous, but that intolerance was no laughing matter.

The first troll-bot to enter the big league was Dunold Tramp, a bombastic AI Twitter bot that quickly gathered a million followers, and plenty of Bitcoin donations. It seems there are plenty of voyeuristic haters out in the wild, who are only too glad to have an anonymous mouthpiece. It was quickly followed by two opposing racist bots spawned in South Africa, which have proved impossible to block, and impossible to prosecute for hate crimes, being non-human.

After Brazil’s failed attempt to block WhatsApp, governments are increasingly wary of authoritarian censorship of ‘free’ speech. No-one wants to be called “North Korea” or accused of supporting the Great Firewall.

Some commentators have welcomed the new fad for hater-bots, saying it helps to expose what some people are really thinking and saying in private; better to know the devils among you, than live in a fool’s paradise of blissful ignorance.