Andrej Karpathy, a co-founder of OpenAI and former Tesla artificial intelligence executive, has joined Anthropic, adding one of the industry’s best-known researchers to the company behind the Claude chatbot.

Karpathy said Tuesday that he had joined Anthropic and was returning to research and development. “The next few years at the frontier of LLMs will be especially formative,” he wrote in a post on X, referring to large language models.

Anthropic said Karpathy has started this week on its pretraining team, according to Reuters. The group works on the large-scale training process that gives Claude its core knowledge and capabilities. Karpathy will report to Nick Joseph, Anthropic’s head of pretraining.

Joseph said Karpathy would build a team focused on using Claude to accelerate pretraining research, according to a post on X. That makes the move more than a senior hire. It places Karpathy in one of the most expensive and strategically important parts of frontier AI development: improving how the next generation of models is trained.

Karpathy’s move comes as Anthropic and OpenAI compete for technical talent, enterprise customers and investor attention. Anthropic, founded by former OpenAI employees, has become one of OpenAI’s main rivals in the market for large language models. Its Claude models are used by consumers, developers and businesses for writing, coding, research and workflow automation.

Karpathy was part of OpenAI’s founding team before joining Tesla in 2017, where he led work on computer vision for Autopilot. He left Tesla in 2022 and later returned briefly to OpenAI before starting Eureka Labs, an AI education company.

His hiring gives Anthropic a researcher with experience in both frontier AI labs and large-scale commercial deployment. At Tesla, Karpathy worked on systems connected to self-driving technology. At OpenAI, he was part of the early group that helped establish the company before it became one of the most valuable private technology firms in the world.

The hire also shows the importance of pretraining in the current AI market. Pretraining is the stage in which large models are trained on vast amounts of data before being adapted for specific tasks or products. Improvements at that stage can affect model quality, cost, speed and reliability.

For AI companies, the cost of building more capable systems is rising. Labs need researchers, data, chips, cloud infrastructure and model-evaluation systems. The companies that improve training efficiency may be able to reduce costs or increase performance without relying only on larger compute budgets.

That is why senior research hires can carry commercial weight. In the AI sector, a small group of researchers and engineers can influence model design, training strategy and product direction. Talent has become one of the key constraints in the race to build more capable systems.

Karpathy’s arrival follows other senior moves at Anthropic. TechCrunch reported that the company also hired Chris Rohlf, a cybersecurity veteran who previously worked at Meta and Yahoo’s security team, to join its frontier red team. That group stress-tests advanced models against serious threats.

Anthropic has also attracted attention for recent hiring from companies linked to Elon Musk. Reuters reported that Karpathy’s move follows a broader trend of high-level transitions in the AI sector, including several former OpenAI executives leaving the company.

The move comes shortly after a federal jury ruled against Musk in his lawsuit against OpenAI and Chief Executive Sam Altman. Reuters reported that the jury dismissed Musk’s lawsuit on statute-of-limitations grounds, easing a legal overhang for OpenAI while leaving questions about the company’s governance exposed during trial proceedings.

Karpathy’s previous work at OpenAI and Tesla drew attention during the Musk-Altman dispute because it highlighted the movement of technical talent between the companies that shaped the early AI sector. Musk helped launch OpenAI, later left its board, and went on to build xAI while continuing to compete with OpenAI and other AI firms.

For Anthropic, Karpathy’s hire strengthens a research organization already positioned around model safety, enterprise use and frontier development. The company has promoted Claude as a system for coding, writing, analysis and workplace productivity, while also emphasizing safety testing and model behavior.

For OpenAI, the move adds to a broader industry narrative about the movement of senior AI researchers among rival labs. OpenAI remains one of the leading companies in generative AI, but competitors including Anthropic, Google, xAI and Meta are investing heavily in models, infrastructure and hiring.

The talent competition has expanded beyond research labs. AI companies are recruiting people with experience in infrastructure, cybersecurity, enterprise sales, public policy, product design and safety evaluation. The sector is no longer competing only to build models. It is competing to turn those models into products, platforms and business infrastructure.

Karpathy’s role suggests Anthropic is putting more attention on using AI systems to improve AI development itself. If Claude can help researchers accelerate pretraining work, the company could gain an advantage in model iteration. The approach also reflects a broader industry push toward AI-assisted research and engineering.

The commercial stakes are high. Better pretraining methods could help companies build more capable models with fewer wasted training runs. They could also improve reliability, reduce costs and shorten development cycles. For enterprise customers, those gains may eventually show up as better coding tools, stronger reasoning systems and more useful workplace assistants.

Karpathy said in his post that he remains interested in education and plans to return to that work in time. For now, his move signals a return to frontier AI research at a moment when the field is becoming more competitive and more capital intensive.

The hiring gives Anthropic another visible name in a market where reputation matters. AI buyers and investors often look at research teams as a sign of long-term capability. The companies with the strongest teams are more likely to attract capital, customers and additional talent.

Anthropic’s challenge will be turning that research strength into durable business gains. The company must compete with OpenAI’s brand, Google’s infrastructure, Meta’s open-source strategy and xAI’s ambitions. Karpathy’s arrival does not settle that contest, but it gives Anthropic another senior figure in one of the industry’s most important technical areas.