The company says its latest models can convert aging code into C++, Java, Kotlin, Python, and Rust without changing behavior.
OpenAI said on October 19 that its latest models show substantial improvements at migrating software written in older programming languages into modern code. The company pointed to COBOL as a prime candidate and said the models can produce working C++, Java, Kotlin, Python, and Rust without altering what the original software does.
COBOL dates to 1959 and still runs a large share of critical infrastructure. IBM estimates put COBOL behind more than 40 percent of online banking systems, 80 percent of in-person credit card transactions, and 95 percent of ATM transactions. Government agencies including the IRS and Social Security rely on it as well.
Replacing legacy code is risky because the systems are too important to break and many programmers who understand them have retired. OpenAI acknowledges that a migration counts as successful only if all tests pass, which sets a high bar for enterprise teams weighing a conversion of production systems.
One rough estimate puts the annual cost of COBOL inefficiencies to US GDP at around $105 billion, which explains the interest in automation. Builders should watch for independent test results and for whether OpenAI or rivals publish migration benchmarks covering Fortran and other legacy languages.
What matters
- OpenAI reported substantial improvements in its models’ ability to migrate legacy code to modern languages.
- COBOL still underpins 95 percent of ATM transactions and more than 40 percent of online banking systems.
- Watch whether OpenAI publishes test results showing migrated systems pass all existing validation suites.
Why it matters
Watch whether OpenAI publishes test results showing migrated systems pass all existing validation suites.
This GenAI News article was prepared in original wording using reporting and materials published by The Verge. Source reference: https://www.theverge.com/ai-artificial-intelligence/1018434/cobol-migration-legacy-code-openai-anthropic.
Drafted by the GenAI News review pipeline.