Bill Foster proposes open-source tools to help regulators track AI financial risks
The discussion draft would test faster reporting and risk monitoring without using actual customer accounts or requiring regulators to adopt the resulting software.
Loading page…
The discussion draft would test faster reporting and risk monitoring without using actual customer accounts or requiring regulators to adopt the resulting software.
Listen to this story
The draft would have the Federal Reserve coordinate a public, open-source software study covering regulatory reporting and financial-risk tools, including AI-risk monitoring. Rep. Bill Foster released it on November 5, 2026, but it is only a proposal: the study would use synthetic or legally permitted historical data, and regulators would take no supervisory action or be required to adopt its software. If enacted, the models would be published within a year, with a report to Congress due within two years.
The proposed software spans banking records, regulatory reporting, stress testing, anti-money-laundering checks and fraud screening.
The Fed would seek technical contributions from financial institutions, technology providers, open-source groups and researchers.
Foster’s office argues shared tools could reduce compliance and development costs for community banks and credit unions.
Rep. Bill Foster released a discussion draft on November 5, 2026, that would put the Federal Reserve in charge of testing open-source software for faster financial oversight. His office says AI-powered services and around-the-clock payments may spread financial stress more quickly, increasing the need to spot emerging risks.
The Real-Time Regulatory Study Act of 2026 would direct the Fed’s Board to work with other financial regulators on demonstration software, then release it publicly. It remains a legislative proposal, not an enacted requirement or an operating monitoring system.
More timely information and automated tools could help institutions and supervisors respond faster, Foster’s office says. The intended benefits include better access to emergency funding and preparation to wind down failing banks.
The proposed project would examine whether regulators can obtain real-time information about the condition of financial institutions nationwide. Its purpose also includes making regulatory obligations easier and less expensive to meet.
Cybersecurity is another part of Foster’s case. His office says AI is creating new threats and argues that open-source software could offer stronger protection through security built in from the start and maintenance shared among contributors.
A Foster spokesperson told Roll Call that the software would provide standard reporting formats, along with tools to manage and transmit information. The intention is to reduce manual compliance work at smaller financial institutions.
The draft’s scope extends beyond submitting reports. According to the bill summary described by Roll Call, software would be developed for:
The Fed would seek technical contributions from financial institutions, technology providers, open-source organizations and researchers. Foster’s office says shared software also has the potential to lower development costs for community banks and credit unions and expand their access to advanced risk-management tools.
Testing would use synthetic—artificially generated—or legally permitted historical data rather than actual customer accounts, according to the summary. Regulators would not take financial or supervisory actions during the study.
If enacted, the measure would require publication of all software models within one year. A final report would be due to Congress within two years, assessing feasibility, costs and savings, competition and other criteria. Those deadlines would run from enactment, not from Foster’s release of the discussion draft.
Loading discussion...
Join the conversation
Explain what transparency would—and would not—tell you about its safety.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.