Anthropic has disclosed that its artificial intelligence assistant Claude generated more than 90% of contributions to the company's software codebase in September 2026, highlighting the growing use of AI-powered coding tools within its engineering operations.
The disclosure was made in Anthropic's registration statement filed with the US Securities and Exchange Commission (SEC) on October 26, as the company prepares for a proposed initial public offering (IPO) on the Nasdaq stock exchange under the ticker symbol ANT.
According to the filing, Claude has become a central part of Anthropic's software development process, with engineers increasingly responsible for directing, supervising and reviewing AI-generated code rather than writing individual lines manually.
The company reported that the average daily volume of code contributed per engineer during the three months ended September 30, 2026, was more than 45 times the average recorded in 2024.
Anthropic attributed the increase partly to the integration of Claude into its internal workflows, allowing employees to undertake development tasks at a greater scale.
Beyond software coding, the company said its AI systems are being used to support research, model development, experimentation and evaluation. Claude is also involved in generating training data, conducting automated assessments and assisting with safety-related testing.
The filing describes a development process in which Anthropic uses its existing models to support the creation and improvement of subsequent versions.
The company stated that Claude has become an important contributor to its research and engineering progress, including improvements intended to increase the efficiency of its computing infrastructure.
Anthropic's disclosures also provide details about changes in its workforce. Its employee count increased from approximately 940 at the end of December 2024 to around 5,300 by September 30, 2026.
The company indicated that AI-assisted workflows have enabled it to expand its operations while increasing the volume of work completed by individual employees.
Anthropic Chief Executive Officer Dario Amodei had previously predicted that AI systems would account for a substantial proportion of software coding activities. In March 2025, he suggested that AI could generate 90% of code within the following three to six months.
The latest filing provides company-reported figures on the extent of AI-generated code within Anthropic's own development environment.
However, the percentage refers specifically to contributions made to its codebase during September and does not establish that AI independently develops or maintains all of the company's software.
Anthropic also identified risks associated with increasingly capable AI systems, including the possibility of unintended or misaligned behaviour.
The disclosure comes as AI-assisted software development becomes a larger part of enterprise technology operations, with organisations examining how automation affects engineering productivity, human oversight and software quality.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.