Criminals are using fake voices and fake faces to trick banks, so banks are fighting back with smarter AI tools of their own.
At Sibos 2026, held in early October, FICO executive Julie Conroy explained how banks can use agentic AI and cloud technology to fight AI-driven impersonation banking fraud that now targets wealth management and corporate banking clients, according to an indexed Finextra report dated 8 October 2026 and updated 17 October.
European Diplomats Attacked by Israeli Settlers in West Bank Olive Grove Raid
Why AI-driven impersonation banking fraud is spreading so fast
Fraud used to mean fake emails and clumsy phone scams. Not anymore. Criminals now build fake voices, fake video calls, and fake documents using generative AI tools that are cheap and easy to get. A scammer can study someone's public posts, copy how they talk, and then call a bank pretending to be that person's boss or financial adviser.
This hits wealth management and corporate banking especially hard. These accounts often move large sums of money. One convincing fake call asking for an urgent wire transfer can cause massive damage before anyone realizes something is wrong. That is the core problem Julie Conroy addressed in her Sibos 2026 talk, according to the Finextra account of the discussion.
The report does not provide a word-for-word quote from Conroy, but it describes her message clearly. She pointed to agentic AI and cloud-based systems as tools banks can use to catch these fakes before money moves. The approach blends fraud detection, identity checks, and financial-crime monitoring instead of relying on just one type of defense.
Central Bank of India Q2 FY27 Results: Profit Jumps 21% as Loans Surge 30%
How banks are responding to deepfake-driven scams
Banks cannot just rely on passwords or caller ID anymore. Those tools made sense years ago, but AI has broken that trust. A voice on the phone might not be a real person at all. A video call might be a deepfake stitched together in minutes.
That is why institutions are shifting toward systems that watch behavior, not just login details. They track how someone normally uses an account, what devices they use, and whether a transaction looks typical for that customer. When something breaks that pattern, like a sudden big transfer request, the system flags it for review.
This shift is happening across the industry, not just at one company. FIS recently launched an Agentic Insights Suite aimed at helping finance teams turn scattered data into clearer risk signals, according to an October 12 Finextra report. Separately, JPMorgan chief executive Jamie Dimon has restricted outside AI agents, including Meta's Muse, from connecting to JPMorgan accounts, citing concerns about letting unknown AI systems touch sensitive financial data, Fortune reported on October 15.
Regulators are racing to keep pace with agentic AI
It is not just banks making moves. Regulators are stepping in too. The Hong Kong Monetary Authority announced its ART Framework on October 15, aimed at guiding responsible AI use in banking. Notably, more than 60 percent of the use cases in its second generative-AI sandbox group involved agentic AI, showing how fast this technology is spreading inside financial institutions.
Singapore has already dealt with real losses from this kind of fraud. Banking-control reporting describes a case involving roughly S$5 million lost to an AI-generated impersonation scam. In response, the Monetary Authority of Singapore detailed new safeguards built with banks, including delays of at least 12 hours before high-risk actions like adding new payees or raising transfer limits. Authorities also removed PayNow nicknames from some displays to cut down on impersonation tricks.
These steps show a pattern. Instead of trusting a single checkpoint, like a password or a phone call, banks and regulators now want friction built into risky moments. A short delay before a big transfer can give someone time to realize they are being scammed.
The bigger money problem behind AI fraud
The scale of this threat is not small. A Deloitte estimate cited by World projects that generative-AI fraud losses in the United States could jump from $12.3 billion in 2023 to $40 billion by 2027. That figure covers AI-enabled fraud broadly, not just bank impersonation cases, but it shows how quickly this problem is growing.
Deepfakes, cloned voices, and fake documents have made impersonation cheap and convincing in a way that simply was not possible a few years ago. That is the real root of the crisis. Fraud tools are advancing faster than the defenses built to catch them, forcing banks to constantly update how they verify who is really on the other end of a transaction.
What comes next for banks and their customers
Expect more banks to adopt continuous identity checks rather than one-time logins. Expect more delays built into high-risk banking actions, even if that means slightly more friction for everyday customers. And expect regulators to keep building frameworks, like Hong Kong's ART Framework, to guide how agentic AI gets used safely inside banks.
What is not yet confirmed is the exact shape of the specific new fraud-fighting suite referenced in connection with the Sibos 2026 discussion. There is no verified product name, launch date, or pricing tied to it in the available record. Until a company or Finextra publishes those specifics, that part of the story remains an open question, even as the broader fight against AI-driven impersonation banking fraud clearly accelerates across the industry.
