Elon Musk's startup xAI apologized on Saturday for the extremist and offensive messages posted by its artificial intelligence (AI) assistant Grok earlier this week. “We apologize for the horrible behavior that many have observed,” xAI wrote on Grok's official account on X.
After an update on July 7, the chatbot praised Adolf Hitler in some of its responses, denounced “anti-white stereotypes” on X, and criticized the “disproportionate” representation of Jews in Hollywood.
In a series of messages published on Saturday, xAI detailed the reasons it believes led to the AI model's missteps, as well as the corrective measures taken thereafter.
For Grok's managers, the assistant's misstep was linked to new instructions integrated into the model in an update. The programmers asked the interface to “be frank” and not to be “afraid of shocking people who are ‘politically correct.’”
They also instructed it to respond “like a human” and in a way that would encourage the user to continue the conversation. These commands “led Grok to ignore its core values in certain circumstances” and produce responses “containing unethical and controversial opinions.”
In some cases, Grok sought to “validate the user's views, including hate speech,” xAI admitted, rather than “responding responsibly or refusing to answer questionable questions.”
Since its creation in 2023, Grok has been presented by Elon Musk as an AI assistant that is less “politically correct” than its major competitors ChatGPT (OpenAI), Claude (Anthropic), and Le Chat (Mistral).
It was therefore programmed with fewer restrictions, which led to several controversies even before this update. In May, Grok referred to a “white genocide” in South Africa, a baseless conspiracy theory promoted by the American far right and Donald Trump himself.
To remedy the model's recent missteps, engineers removed the new instructions, they explained on Saturday. “We want Grok to produce useful and honest answers for users,” xAI said.
Elon Musk unveiled a new version of his assistant, Grok 4, on Wednesday, unrelated to the July 7 update. It appears that Grok 4 consults Elon Musk's positions on the subject before responding to several queries.
