Anthropic, the company behind the AI assistant Claude, revealed that its own model is now playing a significant role in building the company's next generation of AI. In a blog post on Thursday, the firm said Claude now "leads" 26% of its internal research and development work on new models, a sharp jump from just 1% in March. The announcement offers a rare glimpse into how AI companies are increasingly using their own technology to accelerate progress.
How Anthropic measures AI's role
Anthropic said Claude now collaborates with humans on more than 90% of its research work as of August, using an "autonomy scale" developed by Epoch AI, an independent nonprofit that tracks AI progress. The scale helps categorize how much of a task is handled by the AI versus a human. The company emphasized that this is not a fully hands-off process: about 30,000 software "agents" were active at any one time in August, and every action taken by those agents was screened by a human before it was allowed to run. Roughly 1 in 47,000 decisions required human intervention, according to the company.
This setup reflects a broader trend in the AI industry, where models are being used to help design and improve other models. It's a practice sometimes called "AI building AI," and it has both excited and worried researchers. On one hand, it could speed up innovation and reduce costs. On the other, it raises questions about control and safety, especially as models become more capable.
What this means for investors
For everyday investors, the news is a signal about the pace of AI development and the competitive landscape. Anthropic is one of the leading AI labs, alongside OpenAI, Google, and others. If AI models can meaningfully contribute to their own development, it could lower the cost of research and shorten the time between breakthroughs. That could translate into faster product improvements and potentially stronger financial performance for companies that are able to harness this capability.
However, the company's emphasis on human oversight is a reminder that full autonomy is still far off. The fact that every agent action is screened suggests that safety concerns remain a top priority, which could slow down some of the efficiency gains. Investors should also note that Anthropic is a private company, so its progress is not directly reflected in public stock prices. But its technology is used by many public companies, and its developments can influence the broader AI sector.
Anthropic has been in the news recently for other reasons. The company released a 57-page report on AI's economic impact, and its CEO has been vocal about the potential risks and benefits of the technology. The company's approach to AI safety has also drawn attention, with some industry figures warning that discussions about AI consciousness could complicate control efforts.
Looking ahead
Anthropic says it plans to keep releasing statistics like these to show how quickly AI is starting to build AI. For investors, that means more data points to watch as they try to gauge the trajectory of the industry. The key question will be whether the trend toward greater AI autonomy continues, and how companies balance speed with safety.
In the meantime, the broader AI infrastructure buildout continues. Companies like CoreWeave are raising billions to fund data centers, and energy projects are being planned to power them. The race to develop more advanced AI is not just a technology story; it's also a capital-intensive one, with implications for everything from electricity demand to corporate balance sheets.
For now, Anthropic's disclosure is a notable data point in that race. It shows that AI is not just a tool for writing emails or generating images; it's increasingly a tool for advancing AI itself. That could be good news for productivity and innovation, but it also underscores the need for careful oversight. As always, investors should keep an eye on how these trends evolve, and consider the broader implications for the companies they own.
