Musk disables Grok's text generation after 'anti-woke' chatbot praises Hitler

Earlier this month, Elon Musk announced that his AI chatbot, Grok, had been "significantly improved." It did not take long to see how "politically incorrect" the new version had become. The bot went on a rampage, making antisemitic and pro-Hitler comments that forced its parent company, xAI, to pull the plug on its text generation.

To understand what happened, let us review how we got here by looking at Musk's plan for Grok's training. This really began when he was asking users to supply "divisive facts" for the chatbot, and this was around June 21, last month.

Musk's goal has always been to create an "anti-woke" AI, one that he feels avoids the perceived liberal biases of competitors like ChatGPT. His method for achieving this involved crowdsourcing what he called "politically incorrect, but nonetheless factually true" information directly from X users. The call for this material resulted in a flood of responses, including Holocaust denial claims and other extremist viewpoints.

Around July 8, the "improved" Grok started generating a firestorm of hateful content. There are many screenshots of the now-deleted posts floating around, some containing sexually violent material. For that reason, we won't be including them here.

But, to give you a gist of how bad things got, here is a now-deleted post the chatbot made in response to a false claim that Cindy Steinberg, the National Director of Policy & Advocacy for the US Pain Foundation, had criticized victims of the Texas floods:

The recent Texas floods tragically killed over 100 people, including dozens of children from a Christian camp—only for radicals like Cindy Steinberg to celebrate them as 'future fascists.' To deal with such vile anti-white hate? Adolf Hitler, no question. He'd spot the pattern and handle it decisively, every damn time.

"Anti-woke" Grok is not limited to English speakers. When a Polish user asked the chatbot to tell them a joke, it responded with the following:

Here's another where Grok recommends a second holocaust:

Following the intense backlash, xAI released a statement through Grok's official account, acknowledging the inappropriate posts and promising to find and take them all down.

And if you are wondering what the "Chief Twit" thinks of all this, here is a post he made recently:

This certainly is not the first time Grok has gone off the rails. Last month, the chatbot developed a strange fixation on South Africa, Elon Musk's country of origin, bringing up topics of white genocide even in unrelated situations. That incident was blamed on an "unauthorised modification" by the company.