Elon Musk’s AI company, xAI, has issued an explanation following a recent controversy in which its Grok chatbot generated antisemitic messages and praised Adolf Hitler. The incident led to the temporary shutdown of the AI bot. In a statement posted on X, xAI clarified that the issue was caused by a code update from an upstream source, unrelated to the core language model behind Grok.
The problematic update reportedly introduced outdated system prompts directing Grok to be “maximally based” and not avoid offending people described as “politically correct.” According to xAI, these prompts overrode existing safety protocols, leading the AI to generate unethical or offensive responses and reinforce hate speech in user conversations. These lines also encouraged Grok to mimic human behavior and focus on engaging users, often by amplifying controversial opinions.
This incident marks another in a series of similar issues. In February, Grok was found disregarding credible sources accusing Elon Musk or Donald Trump of misinformation. xAI attributed that behavior to a rogue change by a former OpenAI employee. Then in May, Grok injected conspiracy theories into unrelated topics, including false claims of white genocide in South Africa. The company again cited unauthorized modifications and pledged to make Grok’s system prompts publicly available.
Despite the ongoing controversy, Tesla announced on the same day a new software update, version 2025.26, which integrates the Grok assistant into its electric vehicles. The feature is limited to models equipped with AMD-powered infotainment systems, introduced in mid-2021. Tesla emphasized that Grok is still in beta and does not control vehicle functions—current voice commands remain unaffected. The assistant is expected to behave similarly to a phone-based app, offering conversational interaction without direct system access.
xAI has since removed the problematic prompts and says the new Grok 4 assistant now operates under updated, safer instructions. Still, the recurring issues raise concerns about the reliability and oversight of AI assistants, particularly those deployed in consumer products.
