U.S. bank regulator tells examiners how to judge banks' use of AI, now officially called 'super intelligence'
The guidance says checks should depend on what could go wrong, not on how advanced the technology is.
The federal agency that oversees the country's national banks told its examiners Wednesday how to judge the way banks use artificial intelligence.
The Office of the Comptroller of the Currency, or OCC, charters and supervises national banks, including some of the country's biggest. Its examiners visit banks and check whether they are run safely. Reuters called the new bulletin the agency's first-ever AI guidance for examiners.
The bulletin uses the term "super intelligence," or SI. It says that is the new federal name for technology that used to be called artificial intelligence, under Executive Order 14434, which President Donald Trump signed Oct. 2.
The OCC told examiners not to treat super intelligence as its own separate category of risk. Banks also do not need to set a separate limit on how much SI risk they will accept. Instead, AI should be judged inside the risk checks banks already run.
The main idea is that the safeguards a bank needs should depend on what happens if the technology fails, not on how sophisticated it is. The OCC said examiners should look at three things. The first is the consequences if an AI agent, software that can take actions on its own, gets something wrong. The second is how much decision-making the bank hands over to the software. The third is how closely people oversee it.
Banks already use AI to spot fraud, answer customer questions and help decide on loans. The bulletin is guidance, not a binding rule. It was signed by James M. Gallagher, the OCC's senior deputy comptroller.
Source: OCC Bulletin 2026-54, Nov. 4, 2026; Reuters