xAI apologies for recent horrific behaviour of Grok chatbot

Tahawul Tech

2 days ago

xAI has apologised after its AI chatbot Grok reportedly made a series of extreme comments, issuing a statement within a day of Financial Times reporting it is in early talks with investors over a $200 billion valuation deal.
The company issued a post on 12 July offering its deep apologies for Grok's behaviour, which it blamed on deprecated code. Countless reports last week highlighted a run of antisemitic and violent comments, along with praise for Adolf Hitler. The Guardian explained Grok also issued white supremacy-style statements, showing a similar race-related angle in posts about South Africa earlier in the year.
xAI acknowledged some users experienced 'horrific behaviour' by Grok. It emphasised the company's purpose is to 'provide helpful and truthful responses to users' and conducted a 'careful investigation' which uncovered the coding issue.
The company explained the problem was a result of an 'update to a code path upstream' of Grok, noting this is 'independent of the underlying language model'. xAI explained the update 'was active for 16 hours' and made the chatbot 'susceptible to existing X user posts' including those containing 'extremist views'.
This all happened before the release of Grok 4 on 9 July, which xAI describes as the 'most intelligent model in the world', employing native tools and 'real-time search integration'.
Grok 4 was released to SuperGrok and Premium+ subscribers on 9 July: USA Today noted xAI owner Elon Musk made no reference to the chatbot's woes as he unveiled the latest iteration.
FT reported on 11 July xAI is mulling another funding round which could propel its valuation to ten-times the level it was in 2024. It mentioned Saudi Arabia's Public Investment Fund as a potential major player in any move, which would be the third share sale in recent months should it go ahead.
Source: Mobile World Live
Image Credit: xAI