GitHub is moving Copilot Chat out of beta and into general availability, widening access to its programming-focused chatbot for developers using Microsoft’s Visual Studio Code and Visual Studio.
The launch turns Copilot Chat into a more central part of GitHub’s paid AI coding product. It also raises a familiar question for teams adopting generative AI in software development: how much help can developers safely take from a model that still needs careful human review?
What GitHub Is Making Available
Copilot Chat first appeared earlier this year for organizations subscribed to Copilot for Business. It later reached individual Copilot customers, including those paying $10 per month, in beta. Now GitHub says the chat experience is generally available for all users who qualify through Copilot access.
As of today, Copilot Chat is available in the sidebar of Visual Studio Code and Visual Studio. It is included in GitHub Copilot paid tiers and is also free for verified teachers, students and maintainers of certain open source projects.
The product is meant to work like a coding companion inside the developer environment. Instead of only suggesting code completions, it lets developers ask questions in natural language and receive guidance while they work.
Shuyin Zhao, VP of product management at GitHub, described the broader strategy this way: “As home to the world’s developers, we’ve brought to market what is now the most widely adopted AI developer tool in history. And code complete was just the beginning.”
What Developers Can Ask Copilot Chat To Do
According to GitHub, little has changed about Copilot Chat since its beta period. The system is still powered by GPT-4, OpenAI’s flagship generative AI model, and is fine-tuned for developer scenarios.
Developers can use Copilot Chat to ask for real-time help with programming tasks. The examples given include explaining concepts, detecting vulnerabilities and writing unit tests.
That makes the tool broader than an autocomplete feature. In practice, the pitch is that a developer can stay inside the editor, ask a question about code or a programming concept, and get a response without switching contexts.
The most important use cases described in the source are:
- Natural-language questions about code and programming concepts.
- Guidance on potential vulnerabilities.
- Help generating unit tests.
- Security-related notifications for certain insecure code patterns.
GitHub also points to filters for insecure code patterns. Zhao said those filters can notify Copilot Chat users about vulnerabilities such as hardcoded credentials, SQL injections and path injections.
The Training Data Debate Has Not Gone Away
Copilot Chat’s wider launch does not resolve the long-running dispute over the data used to train generative AI coding systems. Like other generative AI models, GPT-4 was trained on publicly available data. Some of that data is copyrighted or under a restrictive license.
Vendors including GitHub argue that fair use doctrine protects them from copyright claims. That position has not stopped coders from filing class action lawsuits against GitHub, Microsoft and OpenAI. The lawsuits allege open source licensing and IP violations.
One practical question is whether codebase owners will be able to opt out of training. Zhao said there is no new mechanism for that tied to the broader launch of Copilot Chat. Instead, she suggested that codebase owners make repositories private if they want to prevent them from being included in future training sets.
That answer leaves a clear tension for public code. The source notes that there are many reasons to keep copyrighted code public, including crowdsourcing bug hunting. For now, GitHub is not introducing a new opt-out system as part of this release.
Why Human Review Still Matters
Generative AI systems can hallucinate, meaning they can confidently produce information that is not correct. In software development, that risk is especially serious because a plausible-looking answer can still introduce security problems or broken behavior.
The source cites a recent Stanford study finding that developers who use AI assistants to code tend to produce code that is less secure than developers who do not use AI assistants. The article says this is partly because the assistants introduce buggy or deprecated code snippets.
Zhao said GPT-4 performs “better” against hallucinations than the older model that once powered Copilot. She also highlighted exploit-mitigating features, including filters for insecure code patterns.
Even with those improvements, Zhao stressed the importance of close human review of AI-suggested code. That is the practical takeaway for engineering teams: Copilot Chat may speed up certain workflows, but it does not remove responsibility from the developer reading, testing and shipping the code.
“GitHub Copilot is powered by OpenAI’s models, which we’ve found to be the best models for the services we offer today,” Zhao said. “We’re in a really strong position to continue empowering developers with the AI tools they need to build better, more secure software at scale — and to have fun while they’re doing it.”
Competition And Cost Pressure Are Building
The launch also comes as GitHub faces pressure to make Copilot more attractive in a crowded AI coding market. In October, Microsoft CEO Satya Nadella told analysts that Copilot had 1 million paying users and ~37,000 enterprise clients.
At the same time, Copilot appears expensive to run. According to a Wall Street Journal piece cited in the source, Copilot loses an average of $20 a month per user, with some customers costing GitHub as much as $80 a month. The report points to the high price of running the underlying AI models.
The source also notes that GenAI coding startup Kite ran into a similar problem and shut down early last December.
GitHub’s best-resourced rival may be Amazon CodeWhisperer. Amazon made CodeWhisperer free of charge to developers without usage restrictions in April. That same month, Amazon launched CodeWhisperer Professional Tier, adding single sign-on with AWS Identity and Access Management integration and higher limits on scanning for security vulnerabilities.
An enterprise plan for CodeWhisperer launched in September. In early November, Amazon “optimized” CodeWhisperer to provide “enhanced” suggestions for app development on MongoDB, the open source database management program.
Beyond Amazon, Copilot also competes with startups such as Magic, Tabnine, Codegen and Laredo, along with open source models including Meta’s Code Llama and Hugging Face’s and ServiceNow’s StarCoder.
That competitive field explains why general availability matters. Copilot Chat is not just another feature inside an editor. It is part of GitHub’s effort to make Copilot feel more useful, more interactive and more embedded in daily software work while the economics and safety questions around AI coding tools remain unsettled.