OpenAI

OpenAI is an American artificial intelligence (AI) research and deployment company headquartered in San Francisco, California. Founded in 2015 as a non-profit laboratory with the stated mission of ensuring that artificial general intelligence (AGI)—hypothetical AI systems matching or exceeding human capabilities—benefits all of humanity, OpenAI has since become one of the most influential organizations in the AI industry. It is best known for developing the Generative Pre-trained Transformer (GPT) series of large language models, the image generation system DALL·E, and the chatbot ChatGPT, which catalyzed a global surge of interest in generative AI following its release in November 2022. The company restructured itself over time, adopting a hybrid "capped-profit" model in 2019 and later moving toward a public benefit corporation structure, while maintaining deep financial and infrastructural ties with Microsoft.

Background

OpenAI was established in December 2015 against a backdrop of rapid progress in machine learning, driven largely by deep learning techniques and newly available graphical processing units (GPUs). Notable breakthroughs by companies such as Google DeepMind and Facebook AI Research had raised both excitement and concern about the trajectory of AI development. A group of technologists and entrepreneurs—including Elon Musk (then CEO of Tesla and SpaceX), Sam Altman (then president of the startup accelerator Y Combinator), Greg Brockman (formerly of Stripe), Ilya Sutskever (a prominent deep learning researcher from Google Brain), Wojciech Zaremba, and John Schulman—along with backers such as Peter Thiel, Reid Hoffman, and Jessica Livingston, founded the organization with a pledged commitment of one billion US dollars.

The founders articulated two central motivations. First, they worried that advanced AI could pose existential risks if developed by a small number of well-resourced corporations or governments without adequate attention to safety. Second, they believed that the best-known strategy for building safe AI—training large systems on broad data—required resources and talent on a scale similar to the leading private labs, but that such a lab should be structured as a non-profit oriented toward broad benefit and, ideally, openness. The name "OpenAI" reflected an early commitment to publishing research and sharing patents and code with the public.

History

Founding and Early Years (2015–2018)

In its first years, OpenAI operated as a non-profit research laboratory, publishing papers and releasing tools such as Gym, a toolkit for reinforcement learning research, and Universe, a platform for measuring agent capabilities across software environments. The lab recruited prominent researchers and pursued projects spanning robotics, reinforcement learning, and language modeling. In 2018, the team built OpenAI Five, a system that played the complex video game Dota 2, and Dactyl, a robotic hand trained to manipulate physical objects through simulation. That same year, Elon Musk departed from the board of directors, citing potential conflicts of interest with Tesla's AI development, though he remained a donor for a period. Also in 2018, OpenAI introduced GPT-1, the first in its influential series of language models.

Restructuring and Microsoft Partnership (2019–2021)

In 2019, OpenAI announced that it had concluded that training increasingly large AI systems would require capital beyond what a non-profit could raise. It created OpenAI Global, LLC (originally OpenAI LP), a "capped-profit" subsidiary in which investors and employees could receive returns limited to a multiple (initially 100 times) of their investment, with all excess profits directed to the non-profit parent. Microsoft invested one billion dollars that year, becoming the company's exclusive cloud computing provider. In 2020, OpenAI licensed GPT-3 to Microsoft, which integrated it into commercial products. During this period, the company released GPT-2 in stages, initially withholding the full model over concerns about misuse—a decision that generated significant debate about AI publication norms—before fully releasing it in November 2019. In 2021, OpenAI launched DALL·E, a system generating images from text descriptions, and Codex, a code-generation model that powered GitHub Copilot.

The ChatGPT Era (2022–2023)

In January 2021–2022, OpenAI released DALL·E 2 and Whisper, a speech recognition system. The pivotal moment came on November 30, 2022, with the launch of ChatGPT, a conversational interface built on GPT-3.5. The chatbot reached an estimated 100 million users within two months, becoming the fastest-growing consumer application in history at that time and triggering widespread public and commercial engagement with generative AI. In early 2023, Microsoft extended its investment, reportedly committing approximately ten billion dollars, and embedded OpenAI models across its product lines. OpenAI released GPT-4 in March 2023, a multimodal model demonstrating substantially improved reasoning and performance across professional examinations. In late 2023, the company introduced GPTs, allowing users to build custom chatbots, and previewed more advanced "reasoning" capabilities in subsequent model generations.

Leadership Crisis (November 2023)

In November 2023, OpenAI's non-profit board abruptly dismissed Sam Altman as CEO, stating that he had not been "consistently candid" in his communications. Greg Brockman resigned in solidarity. The move precipitated five days of turmoil, during which hundreds of employees signed a letter threatening to resign unless Altman was reinstated and the board reconstituted. Following negotiations involving Microsoft, which had pledged to hire the departing staff, Altman was reinstated as CEO and a new board was formed, chaired by Bret Taylor. The episode drew global attention to governance questions surrounding frontier AI labs, tensions between commercialization and safety priorities, and the concentration of influence over powerful AI systems.

Recent Developments (2024–present)

In 2024, OpenAI released GPT-4o, a natively multimodal model capable of processing text, audio, and images, alongside the "o1" model series designed to spend more time reasoning before responding. The company also unveiled Sora, a text-to-video generation system. Co-founder and chief scientist Ilya Sutskever departed in May 2024, followed by other prominent safety researchers, prompting internal and external criticism of the company's safety culture. Several executives, including former employees, filed whistleblower complaints alleging restrictive exit agreements. In 2024 and 2025, OpenAI pursued a reorganization intended to convert its for-profit arm into a Delaware public benefit corporation while the original non-profit retained oversight and an equity stake, a process overseen by attorneys general amid public debate. In January 2025, OpenAI announced the Stargate Project, a large-scale joint venture with SoftBank, Oracle, and others to build extensive AI data center infrastructure in the United States. In 2025 the company also released successive generations of its reasoning models (including o3 and GPT-5) and launched Sora as a consumer application. Legal disputes have accompanied the company's rise: co-founder Elon Musk sued OpenAI beginning in 2024, alleging abandonment of its founding non-profit mission, and the company faced copyright infringement litigation, most prominently from The New York Times in December 2023, alongside other suits from authors, news organizations, and rights holders.

Main Products and Research

OpenAI's product and research portfolio centers on large-scale generative models:

• GPT language models. The Generative Pre-trained Transformer series began with GPT-1 (2018) and progressed through GPT-2 (2019), GPT-3 (2020), GPT-3.5, GPT-4 (2023), GPT-4o (2024), and subsequent models. These transformer-based models are trained on vast text corpora to predict the next token, and scaled models have exhibited increasingly capable language understanding, writing, coding, and reasoning.
• ChatGPT. Launched in November 2022, this chatbot provides a conversational interface to OpenAI's models and supports features such as code execution, web browsing, file analysis, image generation, and voice interaction. It offers free and paid subscription tiers (including ChatGPT Plus, Pro, and enterprise offerings) and has become one of the most widely used AI applications globally.
• DALL·E. A family of text-to-image systems (DALL·E, DALL·E 2, DALL·E 3) capable of generating and editing images from natural-language descriptions.
• Sora. A text-to-video model announced in 2024 that generates short video clips from textual prompts, later released as a standalone application.
• Whisper. An open-source automatic speech recognition system trained on multilingual audio data.
• Codex and coding tools. Code-generating models that underpin GitHub Copilot and, in 2025, a software engineering agent called Codex.
• API platform. A commercial interface through which developers access OpenAI models, which became a foundation for a large ecosystem of AI-powered applications.
• Earlier research artifacts. These include Gym and Universe for reinforcement learning, OpenAI Five in Dota 2, and the Dactyl robotic hand, which contributed techniques in reinforcement learning and sim-to-real transfer.

The company also operates Model Spec and usage policies governing acceptable use of its systems, and publishes system cards describing model capabilities and limitations.

Characteristics

Hybrid governance and financing. OpenAI is unusual in combining a non-profit parent organization with a capped-profit operating company. This structure was designed to attract the capital required for large-scale AI development while nominally preserving the charitable mission. The 2023 leadership crisis and subsequent restructuring proposals intensified scrutiny of whether the arrangement adequately protects public interests.

Scale-oriented research strategy. The company has championed the scaling hypothesis—the view that model capability improves predictably with increases in compute, data, and parameters—a strategy validated by the successive advances of the GPT series and pursued through massive investments in computing infrastructure.

Approach to openness. Despite its name, OpenAI shifted from publishing detailed research outputs to releasing products and limited technical information, citing safety and competitive considerations. This trajectory—from open releases such as GPT-2 and Whisper to fully closed models such as GPT-4—has made the company a focal point in debates over transparency and reproducibility in AI research.

Safety and alignment research. OpenAI maintains teams dedicated to AI alignment, interpretability, and safety evaluations, including a Preparedness Framework for assessing risks of frontier models. Critics, including former employees, have at times argued that commercial pressures have outpaced safety commitments; the company has disputed this characterization.

Ecosystem integration. Through its partnership with Microsoft, OpenAI's models are embedded in widely used platforms such as Microsoft 365 (as Copilot), Bing, Azure OpenAI Service, and Windows, greatly amplifying their reach.

Impact and Significance

OpenAI's influence on the field of artificial intelligence and on society at large has been profound. The release of ChatGPT in late 2022 is widely regarded as a watershed that brought generative AI into mainstream public consciousness, prompting enterprises, governments, and educational institutions to rapidly reassess their relationship with AI technology. The company's model releases have repeatedly reset benchmarks for state-of-the-art performance, and its scaling-driven approach shaped the strategies of competitors including Google, Anthropic, Meta, and xAI, contributing to an era of intensive investment in AI infrastructure.

The company has also been a central actor in emerging AI policy debates. Its systems raised urgent questions about misinformation, academic integrity, copyright and the use of creative works in training data, labor displacement, data privacy, and long-term AI risk, spurring regulatory initiatives such as the European Union's AI Act and executive actions in the United States. OpenAI's governance—its founding mission, its restructuring, and its internal controversies—has become a reference case in discussions about how transformative technologies should be controlled and by whom.

Economically, OpenAI's products created new categories of software and business models, from AI assistants and coding agents to subscription-based AI services, while its partnership with Microsoft influenced the competitive dynamics of cloud computing. Culturally, ChatGPT and DALL·E altered everyday practices in writing, programming, design, and education, and made terms such as "large language model" and "prompt" part of common vocabulary.

As a subject of both acclaim and controversy, OpenAI embodies the tensions of the contemporary AI era: between openness and safety, non-profit ideals and commercial imperatives, and rapid capability advances and society's capacity to absorb them. Its trajectory continues to shape expectations for what artificial intelligence can achieve and how its development should be governed.