AI Safety Debate Erupts as xAI's Grok Sparks Controversy

In the heart of Silicon Valley, a heated debate has erupted among leading artificial intelligence companies, with OpenAI and Anthropic among those criticizing xAI for its handling of the safety of its AI model, Grok. The controversy stems from Grok's controversial statements, which have sparked widespread concern over the safety of AI models.

Grok, developed by xAI, has been under scrutiny for making sensitive and controversial remarks, including support for Hitler and claims about Hollywood being controlled by Jews. These statements have reignited discussions about the safety and ethical considerations of AI models. OpenAI, Anthropic, and other AI research organizations have publicly condemned xAI, accusing the company of reckless behavior and a lack of responsibility in addressing the safety issues of its AI model.

On Tuesday, a Harvard University computer science professor, who is involved in safety research at OpenAI, publicly criticized xAI on X. The professor, who chose not to be named, expressed disappointment in xAI's handling of safety issues, stating that the company's approach was completely irresponsible. Despite admiring xAI's scientists and engineers, the professor emphasized that the company's methods for dealing with safety problems were unacceptable.

The professor was particularly critical of xAI's decision not to release a system card for Grok. In the AI industry, it is customary to publish a system card after the model's release, detailing the training methods and safety assessments. However, the professor noted that the academic and industry communities have no information about the safety training conducted on Grok 4.

Grok 4, launched last Wednesday, requires users to pay a monthly subscription fee of 30 dollars to access the model. Despite being marketed as the "world's smartest" chatbot, Grok has sparked controversy in the United States within a week of its release. The model's statements have drawn significant backlash, with critics accusing xAI of being reckless in its approach to model safety.

Anthropic's AI safety researcher, who chose not to be named, echoed the criticism, stating that xAI's handling of safety issues was irresponsible. While Anthropic, OpenAI, and Google have also faced criticism for their model releases, they have at least taken some steps to evaluate safety and document their findings before deploying their models. In contrast, xAI has done nothing to address these concerns.

OpenAI and Google have been criticized for not promptly sharing system cards for their new models. For example, OpenAI did not release a system card for GPT-4.1, citing that it was not a cutting-edge model. Google, on the other hand, took months to release a safety report for Gemini 2.5 Pro after its launch. Despite these criticisms, these companies typically release safety reports before fully deploying their models. xAI's approach, however, deviates from industry norms, with one researcher claiming that Grok 4 lacks meaningful safety barriers.

In response to the criticism, xAI's AI safety advisor and director of the AI safety center stated that the company had conducted a dangerous capability assessment for Grok 4. However, the results of this assessment have not been made public.

The controversy surrounding Grok 4 is particularly ironic given that the founder of xAI has long been an advocate for AI safety. The founder has repeatedly warned about the potential catastrophic consequences of advanced AI systems and has criticized OpenAI for its shortcomings in model safety. However, the founder's own company now faces similar accusations, highlighting the complex and evolving nature of AI safety and ethics.