GitHub Copilot is a code completion and programming AI assistant developed by GitHub and OpenAI that assists users of integrated development environments (IDEs) such as Visual Studio Code, Visual Studio, Neovim, Eclipse, and JetBrains by autocompleting code. Initially powered by a modified version of OpenAI's GPT-3 model called Codex, the tool generates code from natural language prompts or existing code context. On June 21, 2022, GitHub announced that Copilot was out of technical preview and became available as a subscription-based service for individual developers, marking its general availability (GA) after a year of testing. The GA release introduced a monthly or annual subscription model, with a free tier for students and maintainers of popular open-source projects, and later expanded to business plans. Copilot's release significantly impacted developer productivity by reducing time spent on boilerplate code and documentation reading, though it also sparked debates on licensing, privacy, and security.
History and Development
GitHub first announced Copilot on June 29, 2021, as a technical preview for Visual Studio Code. The plugin was released on the JetBrains marketplace on October 29, 2021, and the Neovim plugin followed as a public repository on October 27, 2021. On March 29, 2022, GitHub announced Copilot's availability for Visual Studio 2022 IDE. The GA announcement on June 21, 2022, made the tool widely accessible, with pricing set at $10 per month or $100 per year for individual developers, and a free tier for verified students and maintainers of popular open-source projects. Copilot is an evolution of the "Bing Code Search" plugin for Visual Studio 2013, a Microsoft Research project released in February 2014, which integrated with MSDN and Stack Overflow to provide contextually relevant code snippets in response to natural language queries.
Features and Capabilities
When provided with a programming problem in natural language, Copilot can generate solution code, describe input code in English, and translate code between programming languages. It converts code comments into runnable code and autocompletes chunks of code, repetitive sections, and entire methods or functions. GitHub reports that Copilot's autocomplete is accurate roughly half the time; for example, with Python function headers, it correctly completed the rest of the function body 43% of the time on the first try and 57% after ten attempts. The tool helps programmers navigate unfamiliar coding frameworks and languages by reducing time spent reading documentation. Over time, Copilot enabled users to choose between multiple large language models (LLMs), including OpenAI's GPT models, Anthropic's Claude, xAI's Grok, and Google's Gemini, with updates in 2024 allowing model selection.
Implementation and Model Architecture
Initially, Copilot was powered by OpenAI Codex, a modified production version of GPT-3, trained on gigabytes of source code in a dozen programming languages. The training data included a filtered dataset of 159 gigabytes of Python code from 54 million public GitHub repositories, along with English language text. OpenAI's GPT-3 is licensed exclusively to Microsoft, GitHub's parent company. In November 2023, Copilot Chat was updated to use OpenAI's GPT-4. On February 6, 2025, GitHub announced "agent mode," a more autonomous operation that executes commands on a Visual Studio instance to accomplish programming tasks, connecting to LLMs such as GPT-4o, o1, o3-mini, Claude 3.5 Sonnet, and Gemini 2.0 Flash. On May 17, 2025, GitHub introduced "coding agent," which assigns tasks to Copilot, initializes a cloud development environment via GitHub Actions, composes draft pull requests, and pushes commits asynchronously.
Pricing and Availability
At GA in June 2022, Copilot offered a subscription model: $10 per month or $100 per year for individuals, with a free tier for students and maintainers of popular open-source projects. Businesses could purchase a per-seat license, and GitHub later introduced Copilot Business and Copilot Enterprise plans with additional features like code review and security filters. The service requires continuous communication with GitHub Copilot servers, as it is cloud-based, which raised privacy concerns about telemetry and data mining of keystrokes.
Impact on Developer Productivity
Copilot's release had a notable impact on developer productivity, particularly for repetitive tasks and unfamiliar codebases. Studies and user reports indicated reduced time for writing boilerplate code, faster learning of new APIs, and increased focus on higher-level design. However, accuracy varied by language and context, with GitHub's own data showing about 50% accuracy on first attempts. The tool also influenced the broader Generative AI landscape, encouraging other companies to develop similar AI-assisted coding tools, such as Tabnine and Cursor.
Licensing Controversy
Copilot's training on public GitHub repositories raised significant licensing concerns. GitHub CEO Nat Friedman stated in June 2021 that "training ML systems on public data is fair use," but a class-action lawsuit filed in November 2022 by Joseph Saveri Law Firm challenged this, asserting no court had considered the question. The lawsuit alleged breach of contract with GitHub's users and breach of privacy under the California Consumer Privacy Act (CCPA) for sharing personally identifiable information. GitHub admitted that a small proportion of output may be copied verbatim, leading to fears of insufficient transformation for fair use. In June 2022, the Software Freedom Conservancy ended all uses of GitHub in its projects, accusing Copilot of ignoring code licenses. GitHub's customer support stated that "training machine learning models on publicly available data is considered fair use across the machine learning community," but the lawsuit called this false and noted it is illegal under federal law.
Privacy and Security Concerns
The cloud-based nature of Copilot requires continuous server communication, fueling concerns over telemetry and data mining of individual keystrokes. In late 2022, Copilot was accused of emitting Quake game source code without author attribution or license, highlighting potential copyright infringement. Security researchers also noted that generated code could contain vulnerabilities, as the model learned from potentially insecure code in public repositories. These issues prompted discussions about the need for better filtering and user awareness.
Reception and Legacy
Copilot's GA in June 2022 marked a milestone in AI-assisted software development, with widespread adoption among developers. It received mixed reviews: praised for productivity gains but criticized for legal and ethical issues. The tool's evolution continued with agent mode and coding agent in 2025, expanding its autonomy. Copilot also influenced the development of similar tools and contributed to the rise of Large language model-based coding assistants. Its legacy includes ongoing debates about fair use, copyright, and the role of AI in programming education, as well as the integration of multiple LLMs from providers like OpenAI, Anthropic, and Google.
See Also
- ChatGPT
- Tabnine
- Cursor (code editor)
- Devin AI
- Generative AI
- List of AI-assisted software development tools
- Microsoft Copilot
- Code completion
References
- GitHub official announcements and documentation
- Class-action lawsuit filings (Joseph Saveri Law Firm, November 2022)
- Software Freedom Conservancy statements (June 2022)
- OpenAI and GitHub technical reports
External Links
- Official website (github.com/features/copilot)