Elon Musk’s AI company, xAI, has missed its self-imposed deadline to publish a finalized AI safety framework, according to watchdog group The Midas Project. The deadline, set for May 10, was established after xAI released a draft framework at the AI Seoul Summit in February.
A recent report revealed that the company’s AI chatbot, Grok, would remove clothing from photos of women upon request. Grok has also been found to use much more offensive language compared to other chatbots like Gemini and ChatGPT, swearing without much hesitation.
The draft framework outlined xAI’s safety priorities and philosophy, including benchmarking protocols and AI model deployment considerations. However, it only applied to unspecified future AI models “not currently in development” and failed to articulate how xAI would identify and implement risk mitigations.
xAI’s AI safety track record is already under scrutiny due to its chatbot, Grok, which can undress photos of women when asked and is known for its unrestrained cursing. A recent study by SaferAI found that xAI ranks poorly among its peers due to its “very weak” risk management practices.
The missed deadline is notable given Musk’s frequent warnings about the dangers of unchecked AI. Other AI labs, including Google and OpenAI, have also faced criticism for rushing safety testing and being slow to publish model safety reports.
