# OpenAI Models Target Legacy Code Migration in Financial Infrastructure New AI models demonstrate high accuracy in translating aging COBOL systems to modern languages. - **Source:** [Gizmodo](https://gizmodo.com/) - **Published:** 2026-10-20 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/5cbada31-2eb2-4d42-9a95-dcdfff4d0467 ## Legacy Code Vulnerability Critical infrastructure relies heavily on outdated programming languages like COBOL and Fortran. COBOL currently processes 95% of all ATM transactions, despite originating in 1959. As engineers trained in these legacy systems retire, the industry faces a dangerous knowledge gap, leaving essential financial and government services increasingly difficult to maintain or update. ## OpenAI Research Initiative OpenAI recently announced that its latest models, GPT-6 Astra and GPT-5.6 Sol, show substantial improvements in migrating legacy code. Guided by enterprise customer feedback, the company developed specific benchmarks to evaluate how effectively AI can translate old programming languages into modern alternatives like Python, Java, Rust, Kotlin, or C++. ## MultiLangBench Performance Metrics The new MultiLangBench benchmark evaluates an AI model's ability to build SQLite from scratch. GPT-6 achieved an 87% pass rate in this specific test. This performance indicates a significant leap in the model's capacity to handle complex programming tasks, providing a foundation for automating the modernization of aging software architectures. ## Precision in Code Translation Translating COBOL requires maintaining strict business logic, such as specific decimal rounding rules. OpenAI tested its models on migrating COBOL programs without altering original behavior. GPT-6.1 Sol, utilizing high reasoning effort, achieved nearly 100% accuracy, demonstrating that AI can preserve the gestalt message and stylistic requirements of complex legacy financial systems. ## Competitive Landscape Shifts Anthropic has also positioned its Claude Code tool as a solution for modernizing COBOL codebases. Anthropic claims that AI can reduce the modernization timeline from years to quarters. This competitive pressure is significant, as evidenced by IBM shares falling 13% following Anthropic's announcement, marking the company's largest single-day decline since 2000. ## The Modernization Skills Gap Younger programmers increasingly rely on AI assistants like Codex and Claude Code, which are primarily trained on contemporary languages. This creates a disconnect where the workforce lacks deep expertise in legacy systems. By automating the migration process, AI tools aim to bridge this divide, ensuring that critical infrastructure remains functional and secure. ## Operational Risk Mitigation Modernizing legacy systems is historically expensive and time-consuming for financial institutions and government agencies. By utilizing AI to handle the translation of core business logic, organizations can potentially lower operational risks associated with aging code. This shift allows teams to focus on maintaining system integrity rather than struggling with obsolete syntax. ## Future Infrastructure Implications The ability to reliably translate legacy code into modern languages represents a major shift for critical infrastructure management. As AI models continue to improve in reasoning and accuracy, the reliance on human experts for manual code migration will likely decrease, fundamentally changing how financial and government entities maintain their core software foundations. ## Sources covering this story - https://gizmodo.com/openai-says-its-models-can-give-the-financial-sector-a-coding-makeover-2000828398 — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._