OpenAI upgrades ChatGPT capabilities for lesser-used programming languages
The AI firm publishes new evaluations showing improved support for niche coding languages, including COBOL and SQLite, to broaden software engineering utility.
Key points
- OpenAI has published research detailing improvements to ChatGPT’s performance in low-resource programming languages.
- New evaluations cover specific tasks in software engineering, SQLite database operations, and COBOL migration scenarios.
- The updates aim to enhance the reliability of the AI assistant across a broader spectrum of coding environments.
OpenAI has released new findings indicating that its ChatGPT service is delivering improved performance when handling programming languages that receive less attention in artificial intelligence training. According to the company’s own research, these enhancements focus on expanding the tool’s utility for developers who work with niche or legacy systems, rather than just popular modern languages.
The firm highlights specific progress in three distinct areas: general software engineering tasks, interactions with SQLite databases, and the complex process of migrating legacy code from COBOL. These evaluations suggest that the underlying models are becoming more capable of understanding and generating code in environments where training data has traditionally been scarce.
What we know
OpenAI states that its work is centred on advancing coding capabilities in low-resource programming languages. The company has conducted new evaluations to measure performance in these specific domains. These assessments include tests on software engineering workflows, the use of SQLite, and projects involving the migration of COBOL systems. The research aims to demonstrate how ChatGPT can support developers working outside the most common language ecosystems.
Defender's checklist
- Review any automated coding assistance tools currently in use to understand their supported language list.
- Verify that sensitive legacy code, such as COBOL, is not being processed by external AI services without proper sanitisation.
- Monitor logs for any unexpected behaviour when using AI-generated code in production environments.
- Ensure that database queries generated by AI tools are validated for security vulnerabilities before execution.
The background in plain language
Low-resource programming languages are those with limited digital documentation or training data, making them harder for AI to master. Developers often rely on major languages like Python, but legacy systems still use older languages like COBOL. When AI struggles with these, it can produce errors. Teams usually validate AI output rigorously in these cases. Check whether your organisation uses such languages and assess the risk of automated errors in critical systems.
What to do and how to stay safe: OpenAI
- Audit your development workflows to identify where AI tools are used for legacy or niche languages.
- Implement strict code review processes for any output generated by AI in these environments.
- Test AI-generated database queries in isolated environments before applying them to live systems.
- Stay informed about vendor updates regarding language support and known limitations in AI coding assistants.
Step-by-step guide: AI Governance for Small Teams: Practical Controls and Cost
General security guidance from the TechWire Beat newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which programming languages are specifically mentioned in the OpenAI research?
The research highlights evaluations for software engineering tasks, SQLite, and COBOL migration.
Does OpenAI claim ChatGPT is now perfect at these languages?
No, the company states it is "improving" performance and has conducted new evaluations, but does not claim perfection.
Who is the intended audience for these improvements?
The improvements are aimed at developers and software engineers who work with low-resource or legacy programming languages.
