Elon Musk's xAI used Anthropic's Claude to train coding models for months, continuing even after access was revoked. The company is now struggling with a depleted pretraining team and underutilized compute resources.
xAI trained its own coding models on outputs from Anthropic's Claude AI system for an extended period, according to reports. The arrangement continued even after Anthropic terminated access, with xAI reportedly using private accounts and the Blackbox AI service to maintain the pipeline.
The situation reflects broader tensions in the AI sector over model training practices and data usage. Anthropic's decision to cut off xAI signals concern about how its outputs were being leveraged for competitor development.
The incident coincides with significant personnel losses at xAI's pretraining division. The team has shrunk to fewer than five people, with several leads departing the company. These departures suggest internal instability or disagreement over the company's direction and practices.
Meanwhile, Musk's significant compute investments have not translated into internal model development momentum. The hardware infrastructure xAI acquired is now being rented to competitors including Anthropic and Google—an ironic outcome given that xAI was positioned as Musk's answer to perceived AI risks from other organizations.
The xAI project, launched in March 2024, aimed to create an AI system free from what Musk characterized as bias in competitors' models. The company raised $6 billion in funding and secured substantial computing resources. However, the training approach and current trajectory suggest the company faces execution challenges.
The reliance on Claude outputs for training raises questions about model differentiation and independence. xAI's inability to maintain its pretraining team through this period indicates difficulty attracting or retaining talent for the project.
The compute rental arrangement suggests a pivot toward infrastructure monetization rather than internal model development as a near-term focus. For a company explicitly founded to compete with Anthropic, OpenAI, and Google in advanced AI development, this represents a notable shift in resource allocation.
A designer at Jane Street reports using Claude AI more than Figma for design tasks, signaling a potential shift in how design work gets done. The post sparked significant discussion in developer communities.
Sakana AI, a Japanese startup co-founded by Transformer co-author Llion Jones, has launched a research lab focused on recursive self-improvement—AI systems that iteratively enhance themselves. The initiative aims to challenge the compute-intensive arms race dominating frontier AI development at major US labs.
Meta is developing Hatch, its first paid AI product, with pricing reaching up to $200 monthly. The AI agent performs tasks like building tools, scheduling appointments, and sending emails based on natural language requests.
South Korea's stock market is outperforming Japan as the nation's leading companies capitalize on the artificial-intelligence wave. Korean firms are projected to generate more earnings than their Japanese counterparts this year.