企业AI风险与准备度_治理滞后于扩张的警讯_12页_253kb

AI Risk & Readiness in the Enterprise Report Summary (2025)

Key Findings

• Risk Outpaces Governance: AI adoption is widespread, but governance and security lag significantly.
  • 93.2% lack confidence securing AI-driven data.
  • 69.5% place AI-powered data leaks as their top security concern (0.4% choose not).
  • 80.2% unprepared for AI regulations.
  • 47.2% have no AI-specific security controls.
  • Only 35.2% have AI risk monitoring and response capabilities.

AI Risk Awareness vs. Reality

• 5.6% they understand AI model risks but have no visibility at all.
• 39.5% in early-stage AI risk assessment.
• 24% aware but not actively managing AI risks.
• 6.4% have full visibility and control over AI risks.

Top Threats of 2025

1. AI-Powered Data Leaks: 69.5% ranked as biggest security concern; 58.4% concerned about unstructured data exposure.
2. Shadow AI: 48.5% concerned about unauthorized/overlooked AI tools.
3. Compliance: 52.8% concerned about regulatory demands.
4. Model Access Risk: 40.3% worry about improper access by AI models.

Missing AI Security Controls

• 47.2% report having no AI security controls.
• Controls tracked: AI-specific data classification (21.0%), Access control for AI models (30.0%), AI risk monitoring/response (21.0%).

Compliance & Regulation Lag

• 80% are not ready or unclear on regulations (Not prepared: 25.3%; Preparing: 54.9%).

AI TRiSM Confidence

• Only 64.4% somewhat confident AI models aren't using unauthorized/sensitive data; 31.8% lack confidence entirely.

Vertical Industry Specifics

• Financial Services: Heavy regulation yet 62% lack protections.
• Healthcare: 52% cite AI regulations as big hurdle.
• Retail/Consumer Goods: 48% have no visibility into AI using PII.
• Technology & SaaS: 42% have no formal AI risk strategy.

AI Risk Governance Gap

• 21.9% have no clear owner for AI security, compliance, and governance.

Strategic Priorities

• 36.1%: Compliance & Governance
• 30.0%: Visibility into AI Risk
• 15.0%: Preventing AI-driven Data Leaks
• 14.2%: Implementing Security Controls

Recommendations

1. Deploy AI risk monitoring and shadow AI detection.
2. Build secure pipelines.
3. Implement AI-aware classification, remediation, and policy enforcement.
4. Embed AI TRiSM throughout development/deployment.
5. Align governance with regulations and accountability.