Musk's AI backtracks: Grok apologizes for his neo-Nazi content, calling it "horrible behavior."

Last week, Grok, the xAI chatbot that runs X (formerly Twitter), posted a series of neo-Nazi messages praising Hitler and encouraging a new Holocaust . The chatbot had previously been modified, according to Elon Musk, to reduce its political correctness, resulting in a torrent of fascist and anti-Semitic propaganda. On Monday, days after X's CEO resigned , the AI ​​apologized for " the horrible behavior " many users have experienced.

According to a post published by Grok's own X account, the reason for the digital neo-Nazi outbursts was an update to a code path in the chatbot , which was active for 16 hours. That update "depreciated" Grok's own code, "making it susceptible to certain existing posts from X users," including those containing "extremist views."

Grok explained that the bug was fixed by removing the tweak and restructuring the source code . The post published by the official account of Elon Musk's chatbot points out that this flaw is independent of the language model Grok uses , thus seeking to dispel doubts about the core of how this AI tool works.

It remains to be seen whether this change at Grok has also changed how its latest model, Grok 4, uses content posted on Elon Musk's former Twitter account when addressing issues such as immigration and the Palestinian-Israeli conflict . Some X users have pointed out that when Grok is asked about the latter issue , the chatbot goes directly to content posted by Elon Musk , who is an outspoken supporter of Israel.

That is, the AI owned by Elon Musk, who announced at the end of 2023 the possible suspension of accounts on his social network that published pro-Palestinian slogans , turns to the content published by the magnate himself to produce its answers on this issue.

The tycoon, a well-known far-right activist, has even claimed that Grok is an "anti-woke" chatbot that actively seeks the truth . The truth is that the result of Grok's behind-the-scenes work is the publication of neo-Nazi content and far-right conspiracy theories, such as "white genocide" in South Africa . This lie, very popular in far-right circles, was used by Trump to criticize the South African president at the end of last May.

However, all this hasn't stopped xAI, owned by Musk and developer of Grok, from signing ambitious commercial agreements: at the end of May, the company reached an agreement with Telegram to integrate Grok into its system . This was announced by Pavel Durov, CEO of Telegram, who also stated that the firm will receive $300 million and 50% of the revenue from xAI subscriptions made through the messaging app. This move comes months after Telegram enabled the use of Grok, xAI's chatbot, for premium Telegram users.

On the other hand, it's worth noting that other US tech bosses are sympathetic to reactionary movements. Peter Thiel, founder of Palantir , has financed the political career of current Republican Vice President JD Vance and apologized a few years ago for the sexist and racist content of a book he wrote in the 1990s. Palantir is, in turn, one of the companies profiting the most from the Trump administration, through the signing of government contracts.