Karen Hao, a technology reporter who spent weeks embedded at OpenAI's San Francisco offices in 2019, discovered a stark disconnect between the company's public mission and its internal reality. What began as an idealistic nonprofit dedicated to developing artificial general intelligence (AGI), a form of superintelligence that could surpass human intelligence, had transformed into a secretive, paranoid operation driven by what Hao describes as "the ideological pursuit of the machine god".

Hao's unprecedented access to OpenAI came just after the company received a $1 billion investment from Microsoft in July 2019. As a young technology reporter with an engineering background and prior experience at a Silicon Valley startup, she was given rare permission to embed herself in the organization for a profile piece. But from her first days on the premises, something felt wrong. "Right off the bat, I started realising that something was not right," Hao, now 32, explained in a recent interview.

What Did Hao Actually Witness at OpenAI?

The restrictions placed on Hao's access painted a picture of an organization far more secretive than its public messaging suggested. She was chaperoned everywhere within the office. Certain floors and meetings were completely off-limits. Security personnel were given her photograph and instructed to watch for any unauthorized appearances. Most strikingly, employees were warned on the company's internal Slack messaging system to avoid speaking with her beyond "sanctioned conversations".

This atmosphere of control extended to the researchers themselves. As Hao spoke with OpenAI's scientists and engineers, she noticed they were visibly nervous about what they said, constantly worried about violating unspoken rules. The irony was sharp: OpenAI had been founded on the principle of transparency and collaboration, yet the company operated with the security protocols of a classified government facility. The overall environment, Hao observed, was "competitive, secretive and insular".

Hao

What transformed OpenAI from an open research organization into this fortress-like operation was the singular obsession of CEO Sam Altman: achieving AGI before anyone else. This competitive drive extended not just to rival companies like Google, but to entire nations, particularly China. Under Altman's leadership, the company's scientists and researchers, many of them brilliant minds in the field, developed what Hao characterizes as a quasi-religious belief in AGI.

How Did OpenAI's Culture Become So Extreme?

The religious fervor surrounding AGI at OpenAI manifested in ways that seemed almost surreal to outsiders. Several former employees recounted to Hao a retreat held in the Sierra Nevada mountains where senior scientists, dressed in bathrobes, gathered around a firepit at a sprawling lodge. During this retreat, Ilya Sutskever, OpenAI's chief scientist known for his brilliance and eccentric personality, burned an effigy representing AGI. Friends who joined the company described how it was only after leaving that they "came back down to earth".

The paranoia about protecting company secrets reached extraordinary levels. Sutskever once mused to colleagues about what he would do if his hand were cut off and used in a palm scanner to unlock OpenAI's secrets. He proposed building a secure containment facility, a "bunker" within which would sit a computer completely disconnected from any network. Another executive, Dario Amodei, used a disconnected computer to write critical strategy documents, connecting it directly to a printer so he could distribute only physical copies to avoid any digital trace.

What most disillusioned Hao, however, was how this relentless pursuit of superintelligence was reshaping not just OpenAI, but the entire AI industry. The company decided that the best path to AGI was to take large language models (LLMs), which are AI systems trained on vast amounts of text to predict and generate human language, and dramatically scale them up. This meant "pouring ever more data into them and training them on supercomputers larger than anyone has ever built in human history," according to Hao.

Steps to Understanding OpenAI's Shift Toward Commercialization

• The Scale-First Philosophy: OpenAI abandoned the traditional AI research approach of testing hypotheses with small, limited datasets. Instead, the company adopted what Hao calls a "brute force" method, feeding massive amounts of data to AI systems in hopes they would develop general intelligence across any field, rather than solving specific problems like detecting Alzheimer's from brain scans.
• The Financial Imperative: The enormous computing power required for this scaling approach cost far more money than a nonprofit could sustain. OpenAI created a for-profit subsidiary to secure the necessary funding, fundamentally changing the organization's structure and incentives.
• The Industry-Wide Cascade: OpenAI's competitors, including Google and others, raced to develop or upgrade their own large language models, creating what Hao describes as "a collapsing of the entire AI field and the entire industry towards a singular approach that is intellectually extremely lazy".

Hao's observations, detailed in her book "Empire of AI: Inside the reckless race for total domination" published last month, argue that OpenAI sparked a race for technological progress that is "rapacious, extractive and bad for humanity." Her account gained unexpected validation when the Musk-Altman lawsuit played out in a California court earlier this month. Musk had accused Altman of breaching a nonprofit contract by shifting OpenAI toward a commercial enterprise. Though Musk lost the case, the trial revealed internal memos, emails, and text messages that corroborated much of what Hao had discovered during her 2019 embed.

"It was good to see lots of what I had discovered being laid out," Hao said when reflecting on the trial evidence.

Karen Hao, Technology Reporter and Author

Hao emphasizes that the trial itself was somewhat of a distraction from the larger story. What matters, she argues, is understanding how OpenAI arrived at this point and what the industry needs to do next. The company that once promised to develop superintelligence transparently and collaboratively had become something far different: a secretive, commercially driven organization pursuing what its own scientists seemed to view as a kind of technological deity. Whether that transformation was inevitable, necessary, or deeply problematic remains one of the most pressing questions in AI today.