xAI cuts 500 data annotation roles

Grok training operation pivots to specialist expertise

Elon Musk’s AI venture, xAI, has laid off around 500 employees as the company shifts its training strategy for its Grok chatbot towards niche expertise.

xAI has cut approximately 500 employees in a sudden shake-up of its data annotation workforce.

According to reports from Business Insider, staff were informed via an internal email that xAI would be moving away from generalist annotation work and instead prioritising “specialist AI tutors” with backgrounds in fields such as science, coding, finance, law and media. The move represents a major restructuring of the company’s largest unit, which at its peak employed over 1,500 annotators.

Affected employees were told their contracts would be honoured until 30 November, or until their pre-agreed end dates, but system access was revoked immediately, according to reports. The decision triggered rapid changes in the company’s Slack channels, which saw membership drop sharply following the announcement.

Structured training data

Annotation teams are central to the development of Grok, providing the structured training data that underpins the chatbot’s ability to generate responses. However, the company says it intends to expand its specialist tutor team tenfold, suggesting a longer-term bet on domain-specific knowledge over general content moderation and tutoring.

The layoffs follow a week of upheaval within xAI’s training division, which included the departure of several senior managers. Staff were reportedly asked to sit a series of rapid-turnaround assessments on subjects ranging from finance and programming to Grok’s ‘personality’ and internet culture, to evaluate their potential fit for specialist roles.

Oversight of the annotation unit now falls under Diego Pasini, who joined the company in January. Neither Pasini nor xAI have confirmed how many of the laid-off roles will be replaced by new specialist hires.

The restructuring highlights the growing pressure on generative AI firms to improve the quality and safety of their models, particularly as competition intensifies with rivals such as OpenAI and Anthropic.