Companies developing software with AI coding tools such as Cursor and Claude Code are facing a new dilemma.

The pace of code production with AI is outstripping review processes for checking the code that is written, creating what is called a "code overload" or a "deluge of code".

The New York Times recently reported that beyond the sharp rise in the amount of code written by AI, it is also a subtle management issue that AI coding tools work better on laptops than in web-based environments on secure servers run by companies such as Amazon and Microsoft.

Joe Sullivan (조 설리번), an adviser at Silicon Valley VC Costanoa Ventures, said this means more engineers download the entire company code onto laptops, and losing a laptop can lead to security risks. He called it a very dangerous case that no one had thought of even six months ago.

The impact from AI is bigger at companies that allow outside code contributions. Digital whiteboard startup Tldraw provides its technology in the form of licensing, but has made its code public and allowed outside developers to contribute.

In this situation, Tldraw founder Steve Ruiz last fall noticed more people than usual trying to contribute to the company's codebase, and said the behaviour of the new contributors seemed somewhat strange.

Some left the code and walked away at the final stage even though they had nearly finished all work, just before signing paperwork. Others ignored clear instructions or posted updates that poured in like spam.

Ruiz concluded it was highly likely these contributors were not humans but AI bots that were hard to handle, and in January he blocked outside access to Tldraw. Ruiz told the New York Times that if AI bots flooded in, they could have put his team, community and the project’s reputation at risk. He said open-source projects and coding platforms such as GitHub are looking for ways to deal with the new reality.

Moves in the industry to resolve problems from the deluge of code are also taking clearer shape.

Companies using AI coding are looking for experienced experts to manage the enormous volume of AI-generated code, and tech companies selling AI coding tools are accelerating development of AI technology that supports reviews.

The New York Times reported that recruiters are increasing hiring of senior engineers to serve as application security engineers who can find errors in code and manage risks that can arise in software, as the volume of AI-written code surges.

But it is not easy to find people with the experience companies need. Sullivan said there are not enough application security engineers worldwide to meet even the demand from U.S. companies.

In the tech sector, efforts are accelerating to try to solve problems from the deluge of code with AI.

Anthropic and OpenAI recently launched AI-based software review agents, and AI coding tool startup Cursor acquired Graphite, which develops code review bots. Cursor has provided AI-based code reviews through its own Bugbot, and added Graphite’s stacked pull request feature through the acquisition. The feature helps developers handle multiple changes at the same time without waiting for approval of code reviews.