As GPT-5.0 hits the scene, a new Medius survey exposes a growing crisis in finance: nearly one-third of professionals admit they can’t already spot AI-generated fake expense reports. How will they handle even more sophisticated, AI-driven fraud?
A recent survey conducted by the spend management platform Medius highlights escalating challenges for finance professionals in detecting AI-generated expense fraud amid the launch of GPT-5.0.
The study, involving 1,000 finance professionals across the US and UK, reveals that nearly one-third (32%) of respondents feel unable to identify fraudulent expense reports at this point, raising concerns that advanced AI models could further complicate fraud detection.
GPT 5.0 AI model, just released by OpenAI, boasts enhanced understanding and generation capabilities, greater fine-tuning and customization, better integration of text with images, audio, or other data types, and improved handling of long-form content. These improvements enable GPT-5.0 to generate more reliable, versatile, and context-aware AI outputs, but also raise concerns about misuse, such as more convincing AI-generated fraud.
Since the release of GPT-4.0 earlier this year, 30% of finance teams have observed an increase in fake receipts. This trend threatens to intensify as more sophisticated generative AI tools, like GPT-5.0, become widely available.
Medius Chief Product Officer Gary Hall commented, “GPT-5.0 introduces a new level of realism and ease for users, which benefits innovation but also presents challenges for fraud prevention. Legacy finance systems are ill-equipped to handle AI-generated documents that appear authentic, making this a critical issue for finance teams.”
The survey also uncovers significant pressures within finance teams: 42% suspect colleagues of submitting fraudulent expenses, while 34% report being pressured to approve questionable claims. Examples of dubious expenses include luxury items such as diamond rings and cars, alongside unusual claims like fees for Japanese schooling and entertainment venues.
Despite clear communication of expense policies, the survey found a notable gap in adherence, with 66% of professionals believing many employees do not closely follow company guidelines. This disconnect highlights underlying cultural and operational challenges.
Additional pain points include inefficient expense management processes such as receipt chasing, approval delays, and manual data entry. Fraud detection remains a significant concern for a third of respondents, while nearly 30% admit to occasionally bending expense rules themselves.
Hall concluded that existing expense abuse, combined with increasingly sophisticated AI fraud tools, was driving businesses toward a compliance crisis. He suggested that finance teams cannot rely on human effort alone but must adopt intelligent anomaly detection to effectively combat these evolving threats.