The financial services sector has transitioned from debating AI adoption to managing the governance of autonomous systems, according to a new report from the Cloud Security Alliance (CSA). The findings suggest that as AI agents take on more decision-making responsibility, limited visibility into AI-related risks is creating a gap in the industry's ability to investigate and mitigate security incidents.
Cloud Security Alliance Survey on AI Deployment and Visibility
The "State of Cloud and AI for Financial Service 2026" report, commissioned by Anjuna, reveals that 62% of surveyed organizations have already deployed AI agents. Despite this adoption, the report highlights a significant visibility gap: while 20% of respondents reported known AI-security incidents, another 21% were unsure if such incidents had occurred.
The survey, conducted between January 15 and March 1, 2026, gathered 340 responses from global financial services professionals in cloud computing, AI, cybersecurity, compliance, and risk management. The data shows that AI is now largely in production, with 35% of organizations actively implementing AI and 9% in advanced adoption, while 49% are exploring or launching pilot programs. Only 27% of organizations reported no AI agent usage.
AI Agent Autonomy and the Shift Toward Agentic Finance
AI agents have become mainstream across several key functions, with customer service (63%), cybersecurity operations (47%), back-office operations (44%), and fraud detection (41%) listed as the top use cases. Among those using these agents, 93% have granted them some form of autonomy.
The report identifies a trend toward "agentic finance," where AI agents initiate and execute payments on behalf of consumers. While 85% of respondents anticipate the arrival of autonomous AI payments, 65% believe this shift will require a new authorization model to manage the process.
Data Leakage and Cloud Infrastructure Risks
The survey identifies sensitive data leakage through AI interactions as the primary AI security concern for 61% of respondents, a figure that exceeds concerns regarding model attacks or adversarial techniques. This suggests that AI risk is viewed primarily as a data management problem.
Regarding broader infrastructure, cloud services are now nearly universal, with 98.3% of organizations using some form of cloud services and 33% relying primarily or fully on cloud-based architectures. However, human-centric risks remain the top cloud security concerns, specifically third-party risk (55%), misconfiguration (52%), and human error (27%).
Executive Support and Governance Frameworks
The report indicates that 91% of respondents have moderate to strong support from executive leadership, which has facilitated AI deployment. According to Troy Leach, Chief Strategy Officer at the Cloud Security Alliance, the industry is moving toward autonomous AI-driven operations, necessitating a corresponding maturation of identity governance and real-time security controls.
Ayal Yogev, CEO and co-founder of Anjuna, stated that deploying AI at scale in the financial services industry requires responsible deployment. He noted that as agents handle sensitive transactions and data, security, policy enforcement, and accountability must be embedded into the foundation of AI initiatives rather than treated as secondary considerations.
Key Takeaways
- 62% of financial services organizations have deployed AI agents, yet 21% of respondents are unsure if they have experienced AI-security incidents.
- 85% of respondents anticipate autonomous AI payments, with 65% believing this will necessitate a new authorization model.
- Sensitive data leakage is the top AI security concern for 61% of organizations, outweighing concerns over model attacks.
FinanceInsyte's Take
The transition toward "agentic finance" suggests a fundamental shift in how payments and authorizations are handled, but the reported lack of visibility into security incidents is a critical vulnerability. Executives should monitor whether the proposed "new authorization models" can effectively mitigate the data leakage risks that currently dominate the industry's concerns. The reliance on cloud-based architectures means that solving for human error and third-party risk remains as essential as the governance of the AI models themselves.
Source: Businesswire
