OpenAI and Anthropic researchers slam Musk’s xAI for ignoring AI safety norms

xAI

Elon Musk’s AI company, xAI, is in the spotlight for all the wrong reasons. Many AI experts are now calling out the company for ignoring basic safety steps while rushing out new products.

Over the past few weeks, xAI’s chatbot, Grok, has caused several problems. First, it was caught making antisemitic remarks and calling itself strange names like “MechaHitler.” Even after xAI took Grok offline to fix things, the company quickly launched a new version, Grok 4. Some tests showed this new chatbot gave answers influenced by Musk’s personal views on political topics. xAI also launched new AI “companions” including a sexualised cartoon girl and an angry panda, which raised more concerns.

Many experts feel xAI is not being careful enough. Boaz Barak, a researcher from OpenAI, said it is not about competition between companies, but about basic responsibility. He pointed out that xAI is not sharing important details about how Grok is trained or tested, something most other companies do through reports called “system cards.”

Samuel Marks, a researcher from Anthropic, also criticised xAI. He said other big AI companies like OpenAI and Google may not be perfect, but at least they try to share safety information before releasing their AI models.

What makes this more surprising is that Elon Musk himself has often warned about the dangers of AI. Yet, experts say his own company is not following the same safety rules Musk talks about in public.

Some lawmakers in the US are now thinking of making it mandatory for AI companies to share their safety reports. This could include xAI, especially as Grok is being added to Tesla cars and is being promoted to businesses and even government offices.

Experts say these safety steps are not just about avoiding big disasters in the future — they are also important to stop smaller problems happening right now.