LANSING – Introduced by Michigan State University Federal Credit Union (MSUFCU) Deepfake detection using AI The company’s call centers are working to combat the growing risk of fraud associated with synthetic voice attacks. The system was developed in partnership with cybersecurity firm Pindrop and launched in August 2024, and has since delivered tangible results. By September 2025, credit unions must: Approximately $2.57 million in fraud damageAt the same time, operational efficiency and customer satisfaction are improved.
Agency providing services Over 367,000 members The company, which manages more than $8.26 billion in assets, introduced the technology as part of a broader effort to strengthen its anti-fraud capabilities. The implementation comes at a time when deepfake fraud is rapidly increasing across the financial industry.
AI deployment expands across anti-fraud systems
Financial institutions are increasingly deploying artificial intelligence to improve fraud detection across digital channels. AI is used to monitor suspicious activity and support anti-money laundering checks. Identify automated or robot interactions in real time. These systems process large amounts of data and flag anomalies that may indicate fraudulent activity.
The use of AI is expanding beyond banking to other digital areas where trust and security are essential. In e-commerce, Amazon uses AI-powered proactive controls to Block suspected fake listings Before brands have to report. Other online marketplaces are also using machine learning to identify fraudulent sellers, detect refund fraud, and Fighting so-called “friendly fraud”if a customer files an unwarranted chargeback claim. Some of the top-rated casino sites in Hong Kong incorporate AI as well. Detect irregular activity Bonus abuse, account takeover, multiple account fraud, etc.
Pindrop industry data shows that deepfake calls have been recorded. 1300% increase compared to previous year By 2024, approximately 1 in 106 calls will be identified as machine-made.
MSUFCU deploys Pindrop for real-time call risk scoring
Through the partnership between MSUFCU and Pindrop, Passport and protect solutions Within the scope of call center operations. The system evaluates incoming calls and Real-time risk score This is done based on multiple metrics such as voice characteristics and call metadata.
Previously, call center agents relied on: Manual authentication questions To verify customer identity. With the new system, Authentication is performed passively before the call is connected to an agent. This allows staff to immediately access risk insights and tailor their response based on the level of identified threat.
Credit union approved implementation to reduce fraud without increasing friction between members Improve call handling efficiency and improve the overall customer experience. The new system provides greater visibility into fraud detection, allowing institutions to: Track and measure risks associated with incoming calls More precisely.
Measurable impact on fraud reduction and efficiency
Since implementing AI-based call analytics, credit unions have reported:$2.57 million in fraud avoided This equates to a significant reduction in potential losses each month.
After implementation, operational performance improved. Average authentication time per call Reduced from 90 seconds to 45 secondsbecause the system replaced manual verification steps with automated background checks. As a result, call processing within the contact center is faster and more efficient.
Reduced call times and a more streamlined verification process further improved customer experience metrics over the same period. credit union Net Promoter Score increased from 55 to 63 The system went into operation shortly after and remained stable for several months.
Rising Deepfake Threat Reshapes Financial Security Strategies
The introduction of AI fraud detection tools reflects a broader response to the rapid rise in deepfake-related threats. At MSUFCU, Fraud incidents have increased by 38% since 2020. Small organizations are increasingly targeted by attackers because their security infrastructure is less advanced.
Experts point out that Human detection alone is not enough To identify deepfake audio. As AI is becoming hackers’ favorite tool, in controlled scenarios the majority of individuals Unable to distinguish between real and synthetic voice recordings. This restriction has accelerated the introduction of automated detection systems within financial services.
Regulators warn deepfakes are becoming a systemic fraud risk
Regulators are increasingly concerned Increased use of deepfakes Especially in fraud schemes in financial services.
In November 2024, the U.S. Treasury will Financial Crimes Enforcement Network (Finsen) We have issued an alert warning financial institutions about scams involving deepfake media created using generative AI tools. The agency said it witnessed Increase in suspicious activity reports Related to fake identity documents, identity verification bypass attempts, and other scams targeting banks and their customers.
Also provided by FinCEN red flag indicator It reminded institutions of their reporting obligations based on: bank secrecy law.
UNESCO also says about deepfakes: A broader “crisis of knowing”argued that synthetic media not only spread falsehoods, but also undermined public trust in the evidence itself. Industry leaders are now seeing this issue as a new crisis posed by rapid advances in generative AI.
