Since its establishment in 2015 until 2023, OpenAI has once again become the focus of public attention. During this time, it has been in the spotlight for launching the cutting-edge large-scale conversational AI, ChatGPT. However, with the growth of its reputation and influence, the company has also faced frequent security and ethical controversies, some of which have even escalated to legal litigations. Internally, a notable contradiction is that the security team criticizes the company's eagerness to develop AI technology and push for commercialization, believing there is a lack of sufficient attention to safety issues.

Microsoft has been investing in OpenAI since 2019, with commitments totaling $13 billion, pushing OpenAI's valuation to over $80 billion. Behind the company's growth, however, some employees have doubted whether the company could prioritize AI safety and have chosen to leave as a result.

From May 15 to 17, 2023, two leaders of OpenAI's super alignment team, including OpenAI's co-founder and Chief Scientist, Ilya Sutskever, and team lead Jan Leike, left the company. Super alignment refers to the process of guiding and controlling AI through technical means, and according to the company's original intention, it is expected that 20% of its computational resources will be dedicated to this effort.

Upon his departure, Jan Leike expressed on social media that he left due to differences with the company's leadership over the company's core priorities. He believes that OpenAI must become a company that places safety first, and that the company should devote more resources to system security, performance monitoring, and preparing for potential risks.

Sutskever, widely regarded as an AI idealist, initiated a resignation decision against then CEO of OpenAI, Sam Altman, in November 2023, leading to a series of internal power struggles. Sutskever advocates for cautious development of AI and opposes large-scale fundraising, while Altman leans more towards aggressively seeking financial support to accelerate AI research.

After the departure of the two leaders, OpenAI dissolved the super alignment team. Remaining members of the team were reassigned to different positions, and the work on super alignment continued in a more loosely organized internal research group.

Security experts are dissatisfied with Altman's aggressive approach to advancing AI technology and his tendency to accelerate commercialization, criticizing him for not giving proper attention to safety issues. In July 2023, Sutskever and Leike established the super alignment team to ensure AI compliance with safety controls. The establishment of this team was also to demonstrate to the public that the company is responsible for AI safety. In this context, in March 2023, over a thousand scientists, entrepreneurs, and AI practitioners called for all AI labs to immediately halt the development of AI models more advanced than GPT-4.

People have always harbored concerns about the catastrophic consequences that could arise from a loss of control over artificial intelligence. Therefore, Altman actively participated in the official U.S. artificial intelligence regulatory hearings in the first half of 2023 and advocated for AI regulation in public on more than one occasion. To address this issue, the Super Alignment Team was consequently founded. At its inception, OpenAI promised to allocate 20% of its computing resources to support their research work. However, starting from December 2023, the progress of the Super Alignment Team stalled, and they published their last piece of research.

Media outlets revealed that despite the promise, the Super Alignment Team never truly received the 20% of computing resources. In the company's regular resource allocation, the proportion that the Super Alignment Team received was always insufficient, and their research needs were often denied by the management. This lack of resources severely hindered the team's research progress, and it also became a major reason for the team members leaving one after another.

And the departure of Super Alignment team members was not an isolated case. Since 2021, OpenAI has seen frequent changes in several technical experts in the safety domain. What lies behind this is the long-standing point of contention—the dissatisfaction among safety domain experts with Altman's aggressive approach to advancing AI technology and his insufficient attention to safety concerns. Under limited computing power resources, OpenAI seems to have sacrificed the investment originally promised for safety research in order to enhance model performance. Lëck publicly criticized that safety and processes are no longer the primary mission of OpenAI. Compared to ungratifying safety research, glamorous products are their current focus.

A founder of a Chinese artificial intelligence basic software supplier analyzed that OpenAI's position within the industry is not unassailable, and it still faces close competition from giants like Google and Meta. “For Altman, it seems particularly crucial at this stage to fight his way out,” he further explained. OpenAI must now consider commercialization issues, as only breakthroughs in technology can bring confidence to the market and investors. Against this background, the safety issues concerned by the Super Alignment Team have become an obstacle in the evolution of OpenAI's technology and commercialization process.

OpenAI was established in 2015 as a non-profit research institution. In 2019, with Altman's appointment, due to continual financial pressure, OpenAI changed its organizational structure, creating a "capped profit" subsidiary that could make profits, and successfully secured a first-round investment of a billion dollars from Microsoft. Altman's concept was that this new organizational structure would not only satisfy investor expectations for financial returns but would also maintain the company's commitment to its vision of "safe and beneficial AGI." As part of the nonprofit organization, the board of directors was responsible for overseeing the balance between the two.

However, in the process of continuous development, OpenAI's pace has increasingly leaned towards commercialization. When Microsoft's investment was introduced in 2019, the management promised employees that they would not pursue commercialization in the short term and would solely focus on addressing problems with AI technology. However, by 2020, Sam Altman made it clear to employees that OpenAI needed to make money to sustain research. At the beginning of 2023, just two months after the release of ChatGPT, the user count skyrocketed past one hundred million, propelling OpenAI to stardom in the field of artificial intelligence and triggering the race for large model technology. Since then, OpenAI seemed to be unable to slow down, requiring more funds to purchase computing power to continuously iterate more efficient large models.

OpenAI started its latest round of funding in December 2023, and according to reports, if this round goes smoothly, the company's valuation is expected to exceed 100 billion dollars. In the second half of the same year, OpenAI faced competitive pressures from all sides, such as AI startups Anthropic and Mistral AI, which are backed by Google, Amazon, and others, and which have launched upgraded large model products. In response to the situation, OpenAI lifted the registration restrictions for its products on April 1, 2024, aiming to increase user numbers, and to bring a higher valuation and more financing opportunities for the company.

Since 2023, with the acceleration of OpenAI's commercialization, the company has frequently been embroiled in controversies and legal disputes related to safety and ethics. For instance, OpenAI and Microsoft were sued for copyright infringement by The New York Times, which accused them of using vast amounts of news content to train ChatGPT and Copilot, and to respond to user questions based on this information. Tamarack Advisory's analyst Huang Shaoping spoke in January this year that OpenAI and Microsoft are facing ethical challenges caused by AI technology. If the litigation results are unfavorable for the two companies, not only could they face hefty legal compensation, but their further development in the AI field may also be restricted.

On May 13 this year, OpenAI released its new large model, GPT-4o. On the day of the announcement, Altman posted "her" on social media, seemingly hinting that the conversational abilities of GPT-4o could simulate the human-AI romance scenario from the movie "Her," in which the female AI character is voiced by Scarlett Johansson. A voice mode "Sky" of GPT-4o, which sounds remarkably similar to Scarlett Johansson's voice, received widespread public attention. Scarlett Johansson confirmed that OpenAI had attempted to contact her in September of last year and before the release of GPT-4o, wishing to use her voice for an AI robot, but she refused. In response, OpenAI publicly stated that they have stopped using the voice "Sky" in their products to show respect for Scarlett Johansson.

OpenAI continues to spark public discussion due to controversies related to safety and ethics. We invite you to share your views, how do you perceive this series of controversies surrounding OpenAI? Please leave your opinion in the comments section.