Foundation and Early Non-profit Years (2015–2018)

OpenAI was announced on December 11, 2015, as a non-profit artificial intelligence research laboratory ⁹¹⁰. The organization's founding team consisted of Sam Altman, Elon Musk, Greg Brockman, Ilya Sutskever, Wojciech Zaremba, and John Schulman ⁹. Early backers and investors who committed to an initial $1 billion in funding included Altman, Musk, Reid Hoffman, Jessica Livingston, Peter Thiel, and Amazon Web Services ⁹¹⁰. According to its charter, the founding mission was to develop artificial general intelligence (AGI) that is "safe and beneficial" for humanity, prioritizing long-term social impact over financial returns ¹⁹. During its initial years, the organization focused on research and released tools such as OpenAI Gym, a platform for reinforcement learning research, and Universe, a software platform for measuring and training AI intelligence ⁹.

In February 2018, Elon Musk resigned from the board of directors ¹¹. OpenAI stated that Musk's departure was intended to prevent potential future conflicts of interest with his work at Tesla, which was developing autonomous driving technology ¹¹. Following his exit, Musk ceased his financial contributions, and the organization began to seek new methods to secure the computational resources required for large-scale model training ⁶¹¹.

Transition to Capped-Profit Structure (2019)

In March 2019, OpenAI underwent a structural reorganization by creating OpenAI LP, a "capped-profit" entity ¹²¹³. The organization stated this move was prompted by the increasing costs of cloud computing and the necessity of attracting engineering talent with competitive compensation ¹²¹³. Under this model, the original non-profit entity (OpenAI, Inc.) remains the sole controlling body of the capped-profit subsidiary ¹²³⁰. Returns for investors and employees in OpenAI LP are capped at a specific multiple of their initial investment—initially set at 100 times for the first round of investors—with any value generated beyond that cap accruing to the original non-profit ¹².

Industry observers characterized this shift as a departure from the organization's initial commitment to transparency and its original non-profit status ³⁷. OpenAI defended the decision by stating that the capital requirements for achieving AGI could not be met through traditional philanthropic donations alone ¹²¹³.

Microsoft Partnership and Scaling (2019–2023)

In July 2019, OpenAI entered into a long-term partnership with Microsoft, which included a $1 billion investment ¹⁴. This agreement established Microsoft as the exclusive provider of cloud computing services to OpenAI through its Azure platform ¹⁴. In exchange, Microsoft received the rights to commercialize some of OpenAI’s technologies, though the non-profit board retained control over the definition of AGI, which is excluded from Microsoft's commercial licenses ¹¹⁴.

The partnership expanded in January 2023, following the public release of ChatGPT and DALL-E 2, when Microsoft announced a multi-year, multi-billion dollar investment in OpenAI, reported to be $10 billion ³¹⁵. This investment was structured to provide Microsoft with a share of OpenAI LP's profits until the investment was repaid, after which Microsoft would retain a 49% stake in the capped-profit entity, subject to the overall profit cap ⁷¹⁵³³.

2023 Leadership Crisis and Board Reorganization

On November 17, 2023, the board of directors removed Sam Altman as CEO and dismissed Greg Brockman as chairman ¹⁶. The board's official statement alleged that Altman was "not consistently candid in his communications," which hindered the board's ability to exercise its responsibilities ¹⁶. The firing triggered internal instability; within days, over 700 of OpenAI’s approximately 770 employees signed an open letter threatening to resign and join a new Microsoft AI research unit unless the board resigned and reinstated Altman ¹⁷.

Key board members involved in the initial decision included Helen Toner, Tasha McCauley, and Chief Scientist Ilya Sutskever, the latter of whom later expressed regret for his role in the removal ¹⁷. Following pressure from employees and major investors, a deal was reached on November 22, 2023, for Altman to return as CEO ¹⁸.

As part of the resolution, the board underwent a reorganization. Most previous members stepped down and were replaced by a new initial board chaired by Bret Taylor, former co-CEO of Salesforce, and including Larry Summers, former U.S. Treasury Secretary ¹⁸. Adam D’Angelo remained as the only holdover from the previous board. Microsoft also secured a non-voting observer seat on the board, providing the company with visibility into leadership decisions without direct control over governance ¹⁸.

OpenAI's product portfolio is centered on generative artificial intelligence, encompassing text, image, audio, and video modalities. Since 2020, the organization has transitioned from a research-focused entity to a commercial provider of AI services through its proprietary API platform and consumer-facing applications ¹, ⁶.

Large Language Models (GPT Series)

The Generative Pre-trained Transformer (GPT) series represents OpenAI’s core technology. The models are trained on large datasets to predict the next token in a sequence, enabling them to generate human-like text, code, and translations ¹.

• GPT-3 and GPT-3.5: Released in 2020 and 2022 respectively, these models established the foundation for OpenAI's commercial API. GPT-3 was noted for its 175 billion parameters, a significant increase over previous iterations ¹.
• GPT-4: Launched in March 2023, GPT-4 is a large multimodal model capable of accepting both text and image inputs ². OpenAI states that GPT-4 performs at a human level on various professional and academic benchmarks, such as passing a simulated bar exam with a score in the top 10% of test takers ².
• GPT-4o: Introduced in May 2024, the "omni" model (GPT-4o) is designed for native multimodality, processing text, audio, and images in a single neural network ⁷. OpenAI reports that the model achieves significantly lower latency in voice interactions compared to previous versions, approximating human response times ⁷.

Consumer Applications

OpenAI's primary consumer product is ChatGPT, an AI-powered chatbot launched in November 2022. The service reached an estimated 100 million monthly active users within two months of its launch, making it one of the fastest-growing consumer applications in history ³.

• ChatGPT Plus and Team: A subscription-based tier, ChatGPT Plus, was introduced in early 2023, providing users with priority access to new models like GPT-4, higher usage limits, and early access to features such as DALL-E 3 and data analysis tools ⁴. The "Team" plan offers collaborative workspaces and administrative tools for small organizations ⁸.
• Mobile Applications: OpenAI provides official ChatGPT apps for iOS and Android, which include features such as voice-to-text through the Whisper model and synchronization across devices ⁷.
• GPT Store: Launched in January 2024, this platform allows users to share and monetize custom versions of ChatGPT, known as "GPTs," which are tailored for specific tasks like tutoring, creative writing, or technical support ⁸.

Image and Video Generation

OpenAI has developed several models for visual content generation, which have been integrated into its broader product ecosystem.

• DALL-E Series: DALL-E (2021) and its successors, DALL-E 2 (2022) and DALL-E 3 (2023), are text-to-image models. DALL-E 3 is integrated directly into ChatGPT, allowing users to refine images through conversational prompts ⁵. According to OpenAI, DALL-E 3 is designed to follow complex instructions more accurately than its predecessors ⁵.
• Sora: Announced in February 2024, Sora is a text-to-video model capable of generating videos up to one minute in length while maintaining visual quality and adherence to user prompts ⁹. As of mid-2024, Sora remains in a limited release phase for safety testing and feedback from visual artists ⁹.

API and Enterprise Services

OpenAI provides developer access to its models through a proprietary API, which is priced based on the number of tokens processed (input and output) ⁶.

• API Platform: Developers can integrate GPT-4, GPT-3.5 Turbo, and specialized models for embeddings, moderation, and speech-to-text into their own applications ⁶. OpenAI also offers "Fine-Tuning" services, allowing organizations to customize models with their own specific datasets ⁶.
• OpenAI Enterprise: Launched in August 2023, this tier provides enterprise-grade security, privacy, and higher-speed access to GPT-4 ¹⁰. OpenAI asserts that data submitted to the Enterprise tier is not used to train its models, addressing corporate data sovereignty concerns ¹⁰.
• Microsoft Azure Integration: Through a strategic partnership, OpenAI models are available via the Azure OpenAI Service, providing Microsoft’s corporate clients with cloud-based AI infrastructure and integration with the Microsoft 365 suite ¹¹.

Specialized Models and Research Tools

Beyond its primary LLMs, OpenAI maintains several specialized tools:

• Whisper: An open-source automatic speech recognition (ASR) system trained on 680,000 hours of multilingual and multitask supervised data from the web ¹².
• CLIP (Contrastive Language-Image Pre-training): A model that learns visual concepts from natural language supervision, used to power the understanding of images in relation to text ⁵.

OpenAI operates under a unique governance model where a non-profit entity, the OpenAI Foundation, maintains ultimate control over its commercial operations ¹. This structure was significantly updated on October 28, 2025, following a recapitalization that converted the organization's for-profit subsidiary into a Public Benefit Corporation (PBC) known as OpenAI Group PBC ¹, ³. As a PBC, the entity is legally required to advance its mission of developing safe artificial general intelligence (AGI) while considering the interests of all stakeholders, rather than prioritizing shareholder returns alone ¹, ².

Board and Leadership

The OpenAI Foundation is governed by a board of directors that oversees both the non-profit and the PBC ¹. As of late 2025, the board is chaired by Bret Taylor and includes independent directors Adam D’Angelo, Dr. Sue Desmond-Hellmann, Dr. Zico Kolter, Retired U.S. Army General Paul M. Nakasone, Adebayo Ogunlesi, and Nicole Seligman, alongside CEO Sam Altman ¹. To maintain safety oversight, the board manages a Safety and Security Committee (SSC), currently chaired by Dr. Kolter ¹. While most directors serve on both the Foundation and Group boards, Dr. Kolter serves as a non-voting observer to the Group board to ensure a distinct focus on safety governance ¹.

Ownership and Microsoft Partnership

Following the 2025 restructuring, the OpenAI Foundation holds a 26% equity stake in OpenAI Group PBC, valued at approximately $130 billion at the time of recapitalization ¹, ³. The Foundation also holds warrants that grant it additional shares if the company’s valuation increases more than tenfold within 15 years ¹. Microsoft remains the organization's largest individual investor and strategic partner, having committed a total of $13 billion ⁴. Although early reports often cited a 49% profit-sharing arrangement, Microsoft's post-restructuring interest was formalized as a 27% equity stake in the PBC ⁴.

Microsoft maintains a non-voting observer status on the board and serves as OpenAI's exclusive cloud provider ⁴. The partnership is subject to an "AGI clause," which stipulates that if OpenAI’s board determines that AGI has been reached, Microsoft’s commercial rights to that specific technology may be terminated ⁴.

Organizational Culture

OpenAI characterizes its internal culture as mission-driven rather than purely commercial, with Sam Altman stating the organization is "not a normal company" ². The structure is designed to attract and retain technical talent while ensuring that the pursuit of commercial success remains aligned with the non-profit's original goal of ensuring AGI benefits all of humanity ¹, ². The organization is headquartered in San Francisco and has seen significant headcount growth since its 2015 inception ¹.

OpenAI's research and development trajectory is characterized by an early focus on reinforcement learning and robotics, which evolved into a specialized emphasis on large-scale generative models and the empirical study of "Scaling Laws." ¹, ⁵

Scaling Laws and Transformer Research

A core tenet of OpenAI’s research philosophy is the observation that artificial intelligence performance improves predictably in relation to the amount of compute, data, and parameters used during training. In 2020, OpenAI researchers published "Scaling Laws for Neural Language Models," which formalized these relationships and provided a theoretical foundation for the development of increasingly large models. ¹ This approach informed the transition from the original GPT model to GPT-4, as the organization shifted its focus toward the Transformer architecture to exploit these scaling properties. ¹

Alignment and RLHF

OpenAI pioneered the widespread application of Reinforcement Learning from Human Feedback (RLHF) to address the alignment problem—ensuring that AI systems behave according to human intent and safety guidelines. ² While early language models were trained primarily to predict the next token in a sequence, OpenAI’s 2022 research on "InstructGPT" demonstrated that fine-tuning models with human-ranked outputs significantly improved their utility and reduced the generation of toxic or untruthful content. ² This methodology became the technical basis for ChatGPT and subsequent iterations of the GPT family. ²

Multimodality and Robotics

Beyond text, OpenAI has contributed to multimodal AI research, specifically in the intersection of computer vision and natural language processing. In 2021, the organization released CLIP (Contrastive Language-Image Pre-training), a model capable of learning visual concepts from natural language supervision, and DALL-E, a generative model for creating images from text descriptions. ³ Although OpenAI initially maintained a dedicated robotics division—developing systems like a humanoid hand capable of solving a Rubik's Cube—the organization disbanded the team in 2021. ⁵ OpenAI stated that progress in robotics was significantly hindered by a lack of available training data compared to the vast quantities of text available for language models. ⁵

Evolution of Research Transparency

OpenAI’s approach to open-source contributions has shifted significantly since its inception. ⁴ During its early years, the organization released tools such as OpenAI Gym for reinforcement learning and published detailed architectural papers. ⁴ However, starting with the staged release of GPT-2 in 2019, the organization cited concerns regarding the potential for malicious use as a reason for withholding full model weights and training code. ⁴ This trend culminated with the release of GPT-4; the accompanying technical report provided no details regarding the model's size, hardware, or training methods, with the organization citing the "competitive landscape" and "safety implications" as primary factors for its proprietary stance. ⁶

OpenAI's approach to safety and ethics is governed by its primary objective of ensuring that artificial general intelligence (AGI) remains aligned with human values and avoids catastrophic outcomes ¹. The organization utilizes several technical and procedural safeguards, including Reinforcement Learning from Human Feedback (RLHF), which integrates human preferences into the model training process to minimize harmful or biased outputs ⁵. According to OpenAI, this process is essential for making models more controllable and reliable before they are released to the public ¹.

Risk Management and the Preparedness Framework

In late 2023, OpenAI introduced its "Preparedness Framework," a living document designed to identify, track, and mitigate catastrophic risks associated with increasingly capable AI models ². The framework categorizes risks into four primary areas: cybersecurity, chemical, biological, radiological, and nuclear (CBRN) threats, persuasion, and model autonomy ². Under this protocol, OpenAI evaluates models against specific safety thresholds; if a model is assessed to have "high" or "critical" risk in any category and cannot be sufficiently mitigated, the organization commits to pausing its development or deployment ². This framework is overseen by the Safety and Security Committee, which was established in May 2024 to advise the board of directors on critical safety decisions ⁶.

Alignment Research and Organizational Changes

A significant portion of OpenAI's safety efforts was previously focused on "Superalignment," a research initiative launched in July 2023 to solve the technical challenges of controlling superintelligent AI systems ⁴. The initiative was led by co-founder Ilya Sutskever and Jan Leike, who aimed to dedicate 20% of OpenAI's total compute resources to this problem over four years ⁴. However, in May 2024, the Superalignment team was disbanded following the departures of both Sutskever and Leike ⁴. Leike publicly stated that the organization’s safety culture and processes had "taken a backseat to shiny products," and he expressed concerns that the team was not being allocated the promised resources to conduct its work ⁴. OpenAI subsequently restructured its safety operations, integrating the former Superalignment researchers into other product-focused and research teams ⁶.

Ethical Controversies and Data Usage

OpenAI has faced scrutiny regarding the ethical implications of its training data acquisition and model biases. In December 2023, The New York Times filed a lawsuit against OpenAI and Microsoft, alleging that the companies used millions of copyrighted articles without permission to train their models ³. OpenAI has countered that its use of publicly available internet data constitutes "fair use" and is essential for the development of generative AI ¹. Additionally, independent researchers have documented instances of algorithmic bias in GPT models, where the systems may produce content reflecting societal stereotypes or discriminatory viewpoints ⁵. In response, OpenAI states it has implemented system-level filters and "red teaming"—a process where external experts stress-test models for vulnerabilities—to identify and mitigate these behaviors prior to model launch ⁵.

Content Safety and Policy

To manage user interaction, OpenAI enforces usage policies that prohibit the generation of harmful content, such as hate speech, instructions for illegal acts, and sexually explicit material ¹. The organization provides developers and users with a Moderation API to help filter these outputs ⁵. Despite these measures, critics have raised concerns about the environmental ethics of training large-scale models, citing the high energy and water consumption required for data centers, as well as the reliance on low-wage contractors in developing nations for data labeling and content moderation tasks ³.

OpenAI has faced significant legal, ethical, and public scrutiny regarding its data collection practices, its transition from a non-profit to a commercial entity, and its internal safety culture. These challenges involve high-profile lawsuits from media organizations and former founders, as well as broader debates over the definition of "open" artificial intelligence.

Intellectual Property and Media Litigation

In December 2023, The New York Times filed a lawsuit against OpenAI and its partner Microsoft for copyright infringement ¹, ³. The publisher alleged that millions of its articles were used without permission to train Large Language Models (LLMs), which the Times claimed can produce "near-verbatim" reproductions of its content and allow users to circumvent digital paywalls ¹. OpenAI has countered that its training processes are "highly transformative" and protected by the fair use doctrine, as the models learn patterns rather than storing copies of data ².

The litigation has further evolved into a dispute over data governance. In May 2025, a federal magistrate judge ordered OpenAI to preserve ChatGPT output logs for hundreds of millions of users to prevent the potential destruction of evidence ³. OpenAI has resisted this order, characterizing it as an "invasion of user privacy" that conflicts with its legal and ethical commitments to users worldwide ³. The organization maintains that its models are designed with safeguards to avoid replicating original material ².

The Elon Musk Lawsuit

In early 2024, co-founder Elon Musk initiated legal action against OpenAI and its leadership, alleging that the organization breached its founding contract by shifting from a non-profit dedicated to the public good to a "closed-source" commercial entity focused on profit ⁴. Musk, who left the board in 2018, claimed OpenAI has become a de facto subsidiary of Microsoft ⁴. OpenAI has rejected these claims, filing a countersuit in April 2025 that accused Musk of utilizing "bad-faith tactics" to hinder the organization's business and advantage his own AI firm, xAI ⁴, ⁶.

Internal records released by OpenAI suggest that Musk previously supported a for-profit transition in 2017 but departed after his proposals to gain full control or merge the entity with Tesla were declined ⁶. The legal battle involves significant financial stakes, including disputes over OpenAI’s valuation—estimated at approximately $130 billion—and a trial is scheduled for March 2026 ⁵, ⁶.

Internal Stability and Safety Concerns

The organization has encountered criticism regarding its internal safety culture, particularly following the departure of key researchers such as co-founder Ilya Sutskever ⁴. External analysts and former employees have suggested that internal conflicts over control and monetization have occasionally distracted the organization from its mission of making AI safe and equitable ⁴. While OpenAI states its current structure as a Public Benefit Corporation ensures the mission remains primary, critics continue to question the transparency of its "open" nomenclature as its core models remain proprietary ⁴, ⁶.

The introduction of OpenAI's Generative Pre-trained Transformer (GPT) models has resulted in significant discourse regarding automation and its influence on the global labor market. A 2023 study co-authored by researchers from OpenAI and the University of Pennsylvania estimated that approximately 80% of the United States workforce could have at least 10% of their work tasks impacted by the integration of large language models ¹. The research indicated that specialized roles in legal services, financial analysis, and software engineering are more susceptible to task automation than manual labor or manufacturing positions ¹. While OpenAI has characterized these tools as productivity enhancers, external economists have expressed concerns regarding potential wage suppression and the displacement of entry-level workers in information-heavy industries ¹. OpenAI has publicly supported the exploration of Universal Basic Income (UBI) as a potential response to long-term labor displacement, with its leadership participating in large-scale studies on unconditional cash transfers ¹. 1. Labor Market Impact Study 2. Environmental Footprint Analysis 3. Educational Integration Guidelines 4. Global Digital Divide Research

The release of ChatGPT in November 2022 prompted a rapid shift in global education systems. Initially, numerous educational authorities, including school districts in New York and Seattle, implemented bans on the software to prevent academic dishonesty and ensure information accuracy ³. According to UNESCO, the speed of generative AI adoption initially outpaced the development of institutional regulatory frameworks ³. By mid-2023, many institutions began transitioning toward "AI literacy" frameworks, integrating the technology as a pedagogical aid for personalized tutoring and curriculum development while shifting assessment models toward oral examinations and supervised writing ³.

The computational requirements for training and operating OpenAI's large-scale models have raised environmental sustainability concerns. Training GPT-3 is estimated to have consumed 1,287 megawatt-hours (MWh) of electricity, roughly equivalent to the annual consumption of 120 average U.S. households ². In addition to energy consumption, the cooling of data centers used for model inference requires substantial water resources. Research from the University of California, Riverside, indicates that a standard conversation with ChatGPT involving approximately 20 to 50 exchanges consumes the equivalent of a 500ml bottle of water for cooling purposes ².

OpenAI's transition to a commercial provider of AI services has sparked debates regarding the "AI divide," the gap between those with access to advanced technologies and those without. Although OpenAI provides free access to limited versions of its models, the high cost of API credits and subscription tiers for advanced models creates barriers for users in low-income nations ⁴. Furthermore, because training datasets are predominantly composed of English-language text, models often demonstrate lower proficiency and higher hallucination rates in non-Western and low-resource languages ⁴. This linguistic bias potentially marginalizes indigenous knowledge and complicates the application of AI in the Global South ⁴. Following its conversion to a Public Benefit Corporation in 2025, OpenAI is legally required to balance the interests of its shareholders with its mission to ensure that advanced AI benefits all of humanity ¹.

• Google
• GPT-4o
• Meta
• Microsoft
• xAI