OpenAI touts COBOL migration gains as experts warn of legacy-system risks

In a blog post, OpenAI said the models could move COBOL into C++, Java, Kotlin, Python and Rust without changing what the original software does. It said a migration is successful only if all tests pass. OpenAI also reported improvements in models’ ability to build database software in various languages, including COBOL.

COBOL, short for common business-oriented language, was created in 1959 and remains embedded in critical infrastructure. According to IBM, it is the foundation for more than 40 percent of online banking systems, supports 80 percent of in-person credit card transactions and handles 95 percent of ATM transactions. It also underpins government systems including Social Security, the IRS and services run by most states. One rough estimate puts the annual cost of COBOL’s inefficiencies to US GDP at around $105 billion.

Replacing it is difficult. The systems are often too important to risk breaking, and many programmers who understand the code have retired or died. The same challenge applies to other legacy languages such as Fortran, still used in scientific computing, weather forecasting, atmospheric and ocean studies and codes critical to nuclear security. For all their advances in modern software development, AI coding tools have struggled with such legacy systems, though recent developments suggest that may be changing.

In OpenAI’s tests, GPT-6.1 Sol achieved 87 percent behavior preservation accuracy at low reasoning effort—the number of computational steps a model spends analyzing a problem before answering—and 98 percent at higher reasoning. At low reasoning effort, GPT-6 Astra scored around 86 percent, GPT-6 Sol 21 percent and GPT-5.6 Sol 5 percent. OpenAI said the results demonstrated progress toward more accurate, affordable code migration for teams modernizing business software.

Anthropic has also moved into the field. In February it published a step-by-step “code modernization playbook” and pitched its coding agent Claude Code as a way to migrate COBOL into modern languages such as Java or Python. It offered no performance benchmarks for such migrations but said “the economics of COBOL modernization have shifted.” Shares in IBM, which derives part of its revenue helping companies maintain legacy systems, nosedived after the announcement.

OpenAI spokesperson Oscar Haines acknowledged in an email to The Verge that legacy code “is still far from a solved problem.” He said the company is listening to feedback from developers and enterprises and exploring ways to build better tests and improve its models’ understanding of coding languages. Anthropic did not provide comment in time for publication.

Vadim Zaytsev, an associate professor of software evolution at the University of Twente in the Netherlands, said “rewriting code is not the same as replacing a system.” He said the largest system he had seen contained 250 million lines of COBOL. The evaluations shown by OpenAI, he said, do not address the key difficulty: breaking a complex legacy system into smaller units that can be replaced individually.