The digital transformation of financial institutions shows no sign of slowing down. As artificial intelligence becomes embedded at the core of operational processes, cybersecurity risks evolve in step with it. This is the central finding of the Netskope Threat Labs report dedicated to financial services, with data covering February 2025 through February 2026.
Shadow AI Recedes, but It Does Not Disappear
Good news for security executives: the financial sector has made meaningful progress in tackling “ghost AI”—the phenomenon of employees using personal AI tools outside the company’s control. In one year, the use of personal GenAI apps fell from 76% to 36%, while the adoption of organization-managed solutions jumped from 33% to 79%.
ChatGPT remains the most widespread tool, present in 76% of organizations, followed by Google Gemini (68%), steadily increasing. Google NotebookLM quickly established itself with 39% adoption, while AssemblyAI, specializing in transcription, saw explosive growth—from 1% to 37% usage in less than a year.
This move toward controlled environments follows a governance and compliance logic. It does not solve every issue, however: 15% of users continue to switch between personal and professional accounts, indicating that enterprise tools still do not fully meet employees’ expectations for simplicity.
Regulated Data, First Target of Breaches
That’s where the trouble lies. The most sensitive and tightly regulated data remain the prime targets of security incidents. According to the report, they account for 59% of policy violations observed in GenAI applications and 65% in personal applications.
Next come intellectual property (20%), source code (11%), and credentials or API keys (9%). The challenge is further compounded by a factor often overlooked: 97% of users rely indirectly on AI via features integrated into their everyday tools, often without realizing it. This invisible usage makes it even harder for security teams to maintain control.
GitHub and OneDrive, New Vectors for Malware Delivery
Attackers, too, have adapted their methods. Rather than leaning on clearly identifiable malicious channels, they now exploit legitimate cloud platforms to disseminate malware. This tactic lets them blend into ordinary network traffic and bypass some traditional security controls.
GitHub has thus become the leading source of malware distribution targeting the financial sector, affecting 11% of organizations. Microsoft OneDrive follows in second place, impacting 8.2% of entities. Trusted environments, diverted from their legal use, are particularly difficult to filter without risking the blocking of legitimate business activities.
What Financial Institutions Block and What They Recommend
In response to these threats, financial organizations have also tightened their application-blocking policies. ZeroGPT is the most frequently blocked tool (46% of organizations), followed by DeepSeek (44%) and PolitePost (43%). On the monitoring side, Google Drive (40%), ChatGPT (28%), and Gmail (27%) lead the list of applications most tightly controlled.
To strengthen their security posture, Netskope Threat Labs offers several concrete recommendations. First, inspect all HTTP and HTTPS traffic to block malware infiltration, including those traversing legitimate platforms. Next, deploy stringent data loss prevention (DLP) policies, with particular focus on source code and regulated data.
The report also recommends the use of remote browser isolation (RBI) for high-risk sites, including newly registered or recently observed domains. This technology runs the browsing session on a distant server rather than on the user’s device, thereby limiting exposure in the event of malicious content. Finally, a selective application-blocking strategy (allow only those with proven professional use) is viewed as preferable to an overly permissive approach.
Note: The Netskope Threat Labs: Financial Services 2026 report is based on anonymized usage data from a subset of sector clients, collected between February 2025 and February 2026.
