Organizations are moving quickly on generative AI because leaders see it as a practical way to improve productivity, streamline operations, and open up new revenue opportunities.

According to Microsoft’s research with over 400 enterprise IT and security decision makers:

• 95% of respondents said their organization is either already using and/or developing GenAI, or planning to do so.
• Only 5% reported no current use or plans for GenAI.
• 66% of organizations are not just using GenAI apps, but are also planning to develop or are already developing their own GenAI applications.

Among organizations that build GenAI apps, they are typically not stopping at a single project. On average, they are working on or have worked on about 13.9 GenAI apps.

The top reasons companies develop or customize GenAI applications include:

• Driving business innovation (58%) – using GenAI to reimagine products, services, and customer experiences.
• Maintaining control of data (55%) – keeping sensitive data within governed environments.
• Integrating with existing systems (52%) – connecting GenAI into current workflows and platforms.
• Managing cost and scalability (49%) – balancing SaaS usage with in-house capabilities.
• Meeting compliance and regulatory needs (44%) – tailoring solutions to sector-specific rules.

In short, GenAI is being treated as a strategic capability, not a side experiment, which explains both the high adoption rate and the growing portfolio of AI applications inside many organizations.

Security and risk leaders are enthusiastic about AI, but they want to say “yes” in a secure way. The research highlights two broad categories of concern: amplified existing risks and new, emerging risks.

Top concerns when using GenAI

For organizations already using GenAI, leaders are most focused on data and content risks:

• Leakage of sensitive data (63%)
• Sensitive data being overshared (60%) – users seeing data they are not authorized to access.
• Inappropriate use or exposure of personal data (55%)
• Inaccurate insights (43%) – AI-generated outputs that are wrong or misleading.
• Harmful or biased outputs (41%) – content that could damage trust or reputation.

Top concerns when developing or customizing GenAI apps

For organizations building their own GenAI applications, the picture is similar but with added complexity:

• Data leakage and exfiltration (60%)
• Inappropriate use of personal data (50%)
• Violations of regulations (42%)
• Lack of visibility into AI components and vulnerabilities (42%)
• Over-permissioned access granted to AI apps (36%)

Amplified vs. emerging risks

These concerns fall into two groups:

• Amplified risks – familiar issues that grow with AI scale, such as:
  • Data oversharing and breaches
  • Shadow IT and “bring your own AI” (BYOAI) – 78% of AI users are bringing their own tools to work.
  • Misconfigurations and gaps in identity or access controls
• Emerging risks – challenges that are specific to AI, including:
  • Hallucinations (incorrect or fabricated outputs)
  • Harmful or non-compliant content (e.g., deepfakes, misinformation)
  • Prompt injection and adversarial prompts
  • Model theft and supply chain vulnerabilities
  • Training data poisoning
  • Excessive agency – AI systems taking unintended actions

These findings explain why 95% of security and risk leaders say their company needs security measures in place for AI apps (third-party, enterprise-ready, and custom-built) within the next 12–24 months.

The research points to a clear theme: organizations want to embrace AI, but with a structured security approach. A practical path to a “secure yes” to AI includes four key steps:

1. Form a dedicated security team for AI

Create a focused group that brings together security, IT, data, and risk stakeholders. Their role is to:

• Understand how GenAI is being used and developed across the business.
• Identify and prioritize AI-specific risks (data leakage, hallucinations, prompt injection, etc.).
• Set policies and guardrails for AI usage, including BYOAI and third-party tools.

2. Optimize resources to secure GenAI

Align existing security capabilities with AI needs rather than starting from scratch. This typically involves:

• Extending data protection, identity, and access controls to AI workloads.
• Improving visibility into where AI apps, models, and plug-ins are running.
• Addressing shadow IT by giving employees secure, approved AI options.

3. Implement a Zero Trust strategy for AI

Apply Zero Trust principles—“never trust, always verify”—to AI systems:

• Enforce strong identity and access management for AI apps and services.
• Limit permissions and data access for AI components to what is strictly necessary.
• Continuously monitor usage and behavior for anomalies or abuse.

4. Adopt a comprehensive security solution for AI

Finally, organizations are looking at end-to-end security solutions that cover:

• Protection for prompts, responses, training data, and retrieval-augmented generation (RAG) data.
• Controls for models, orchestrators, and plug-ins across the AI stack.
• Support for regulatory compliance, including emerging AI regulations such as the EU AI Act, where violations can reach up to €35 million or 7% of annual turnover.

By following these steps, security and risk leaders can move from blocking or slowing AI initiatives to enabling them, with clearer governance and a more resilient security posture around GenAI.