Can AI help solve the COBOL code crisis?
In brief
- OpenAI reports progress in migrating COBOL code to modern languages with up to 98 % behavior preservation.
- Anthropic releases a modernization guide without publishing tool performance benchmarks.
- Experts warn that COBOL system complexity still makes migration risky despite AI advances.
Critical systems like those of the IRS, banks, or airlines still rely on aging COBOL code that is difficult to modernize without risk. OpenAI claims significant progress with its latest models to migrate this code to modern languages, achieving up to 98 % behavior preservation in internal tests. Anthropic has also released a modernization guide for COBOL code, though without publishing benchmarks.
Experts acknowledge these advances but emphasize that migrating complex systems goes beyond code rewriting. It involves understanding module interactions, data management, and rigorous testing to avoid failures. Current tools still struggle to reliably break down these systems into replaceable units.
Modernizing COBOL systems remains a major challenge, despite growing interest from AI players. Solutions proposed by OpenAI and Anthropic could speed up the process, but their real effectiveness depends on handling the complexity of existing infrastructures and ensuring full compatibility with live systems.
What it changes for you
- Finance & BankingBanks could accelerate the modernization of their aging COBOL systems, reducing maintenance costs and risks, but must rigorously validate migrations to avoid critical failures.
- Public SectorPublic administrations could modernize their critical infrastructures, but migrating systems like those of the IRS or Social Security requires exhaustive testing to ensure service continuity.
This summary gives you the essentials. The full article, its figures and quotes are on AI | The Verge.
Read the original articleAI-generated summary of the source, checked by rules. How we write · Report an error