By: Ben Yi, Grade 10
With AI being one of the biggest markets in the world right now, it’s difficult to think of a world where it won’t be integrated into everyone’s lives within just a few years. Still, however, with how big AI currently is, the business model still runs on a highly inefficient structure that focuses more on the consumer base as opposed to revenue, which can be seen as good for customers hoping to use their products and bad for people within these companies.
One of the main problems is with the massive investments that are put into creating AI models and seeing results of little to no returns on those investments. ChatGPT, for example, is completely free for the average user but despite its paid premium options, still operates under OpenAI’s hybrid structure where a nonprofit foundation controls a for-profit public benefit corporation. More recently, however, OpenAI began testing ads inside ChatGPT’s responses in February 2026, initially rolling them out to US users on the Free and Go tiers as a way to generate more revenue while keeping the platform accessible. Sora, the text-to-video software and app developed by OpenAI, is currently in the process of being shut down, announced on March 24, 2026, partly due to extreme operating costs, low user engagement, and lack of profitability. At its peak, the app had fewer than 500,000 active users while burning through roughly $1 million per day to keep running.
With Anthropic’s development of the AI model named Mythos, people have also become quite afraid of the possibilities of what AI models are currently capable of. Mythos is an advanced, restricted general-purpose AI model developed by Anthropic that has demonstrated extraordinary capability in cybersecurity, more specifically in autonomously detecting and exploiting software vulnerabilities faster than any human could. While Anthropic has not released it to the public, they also launched Project Glasswing alongside it, an initiative aimed at using Mythos to help secure the world’s most critical software before similar capabilities fall into the wrong hands. Those fears were given weight when, on the very day of its announcement, a group of unauthorized users reportedly accessed the model through a connection to a third-party contractor, raising serious concerns about how quickly even restricted AI can escape its intended boundaries.
AI stocks are currently dropping due to investor concerns over the high cost of infrastructure and investments not yielding immediate profits which, if not addressed, could spell trouble for the future business models of AI.
Photo from Scientific American
