Anthropic Relies On Claude To Build Claude. Why That’s Risky

More than 90 percent of Anthropic’s code was written by its own AI tool Claude in September, the company disclosed in its S-1 filing on Monday.

The AI giant is increasingly using its own AI models to research, design and develop subsequent versions of its models— a process it calls “closing the loop.” “Claude itself has become one of the most powerful drivers of our progress,” the company wrote in its S-1.

Today, Claude runs experiments, generates training data and automates evaluations and safety audits. It has also helped make algorithmic improvements that allow Anthropic to use computing resources more efficiently, the company says.

Anthropic, which reported $16.3 billion in revenue and $693 million in operating losses in the first six months of 2026, is slated to go public in the coming weeks. In Q2 2026, its first profitable quarter, it had $974 million in operating income. Today’s S-1 is the first glimpse public investors have gotten into the AI firm’s financials.

It’s not surprising that Anthropic’s 5,300-strong staff relies heavily on Claude. But the company now spends more on internal AI usage than it does on salaries. In April 2026, the value of internal Claude usage exceeded monthly personnel expenses excluding stock compensation, the company disclosed. The value of Anthropic’s internal Claude usage at list prices has grown over 250 times from January 2026 to September 2026, according to the S-1. Thanks to Claude, engineers contributed over 45 times more code per day in Q3 2026 than in 2024.

Claude Code works on complex engineering tasks on its own, with engineers largely supervising agents rather than writing code themselves. Employees across departments like operations, go-to market and research also use a team of agents to automate daily work, allowing smaller teams to do more work and “multiplying the impact of each person,” the company said.

MORE FROM FORBES