It seems all we ever hear in the news nowadays is about artificial intelligence, and nowadays, AI can’t be separated from its spiritual founder, OpenAI, and its service, ChatGPT. This started the AI craze, where artists and software developers had to compete with robots for employment. Like most major technological breakthroughs, the public quickly adopted it before fully understanding the technical side that made it possible.
OpenAI Training Models
Generative pre-trained transformer models like ChatGPT are trained in three key stages:
- Data Collection:
These models digest vast amounts of publicly available text from books, websites, scientific papers, social media, and other sources. This data often includes copyrighted material such as novels, news articles, or code repositories.
- Training:
Learning statistical patterns in the data can begin to predict the next word in a sequence. It does not “store” or “copy” text by specific words but encodes abstract relationships between words, concepts, and styles.
- Generation:
Finally, the model can generate text by sampling from all the patterns it has learned from. Outputs may resemble human writing styles or specific content on which it was trained, but they are synthesized, not direct reproductions.
To be succinct, it synthesizes information in a useful and accessible manner.
The main problem might have already been realized: it needs to ingest mountains of data to create a base, which, in the case of GPT, was used to justify training models on copyrighted material without compensation.
One critic of this approach was Indian American Suchir Balaj, born in California on November 21, 1998. Balaji displayed an uncanny aptitude for technology from childhood. By 11, he was coding with Scratch; by 13, he’d built his computer; by 14, he authored a scientific paper on chip design. At 17, he was recruited by Quora. Later, as a graduate of the University of California, Berkeley, he, a computer science student with a 3.9 GPA, interned at OpenAI and Scale AI.
By 2020, he joined OpenAI full-time and contributed to groundbreaking projects like WebGPT and GPT-4. Balaji’s work on ChatGPT’s training data—scraped from millions of copyrighted books, articles, and code repositories—began to trouble him. By 2024, he would become one of the first OpenAI employees to publicly challenge the company’s ethical compass, a decision that would define his legacy and, many believe, cost him his life.
OpenAI Whistleblower Suchir Balaji and His Crusade
In October 2024, Suchir Balaji penned a viral blog post titled “When Does Generative AI Qualify for Fair Use?”—a meticulous critique of AI’s copyright loopholes. He argued that AI companies like OpenAI were exploiting “fair use” doctrines, training the model using copyrighted information without due process of asking for permission or purchasing licenses. His stark conclusion was that generative AI and GPT often create market substitutes for original works, eroding revenue for creators and publishers.
Essentially, it made competitors for internet services by using the data from those same services. It’s like robbing a store and using the stolen merchandise to start a competing business down the street.
Balaji cited a study showing a 12% traffic decline for Stack Overflow along with a decline in question posting volumes per topic, and lastly that the average account age of a question-asker trended up after the release of ChatGPT, suggesting that newer members are either not joining or are leaving the community. He highlighted how OpenAI’s secretive data practices and reliance on reinforcement learning (RLHF) to reduce “hallucinations” effectively turned AI outputs into low-entropy derivatives of copyrighted works.
At the time, OpenAI faced lawsuits from media giants like The New York Times and The Guardian, who accused the company of “industrial-scale copyright infringement.” Balaji’s technical expertise made him a pivotal witness—and a thorn in OpenAI’s side. When he resigned in August 2024, he left with a warning: “If you believe what I believe, you have to just leave the company.”
On November 26, 2024, Balaji was found dead in his San Francisco apartment at the young age of 26. The San Francisco Police initially ruled it a suicide, citing no evidence of foul play. But inconsistencies quickly surfaced. His family reported he’d been in high spirits days earlier, celebrating his birthday and planning a January visit home. The absence of a suicide note made the whole situation suspicious.
The family commissioned a private autopsy, which allegedly contradicted the official narrative. Dr. Joseph Cohen, the pathologist, reportedly found that the bullet wound in Balaji’s head had missed the brain, and another injury on the other side of the head indicated signs of a struggle, implying the presence of an attacker.
Their claims gained traction when figures like Elon Musk and Congressman Ro Khanna echoed calls for transparency. “This doesn’t add up,” Musk tweeted, amplifying a conspiracy theory that Balaji was silenced to protect OpenAI.
OpenAI, in a statement published after Balaji’s death, said: “We were devastated to learn of this tragic news and have been in touch with Suchir’s family to offer our full support during this difficult time. Our priority is to continue to do everything we can to assist them.”
“We first became aware of his concerns when The New York Times published his comments, and we have no record of further interaction with him. We respect his, and others, right to share views freely.”
The San Francisco Medical Examiner’s Office, however, stood by its suicide ruling, delaying the release of the full autopsy report until February 2025. Meanwhile, the police investigation remains “open and active,” leaving room for speculation about corporate espionage, legal retaliation, or a tragic mental health crisis exacerbated by Balaji’s whistleblower stress.
Suchir Balaji’s death has become a rallying cry in the debate over AI ethics. One truth remains as the world awaits the autopsy report and the outcome of OpenAI’s legal battles. In Balaji’s words: “The gap between possible and impossible is smaller than we think.” For AI, that gap may yet be bridged by accountability—or torn wider by profit. Either way, his voice lingers, a man fighting for ethics in a place where it was a scarce commodity.
If you want to submit your articles and/or research papers, please check the Submissions page.
The views and opinions expressed in this article/paper are the author’s own and do not necessarily reflect the editorial position of Paradigm Shift.
The author is studying Economics at the National University of Science and Technology (NUST) with a keen interest in financial affairs, international relations, and geo-politics.