As Generative AI becomes deeply embedded in enterprise workflows, a new report by Palo Alto Networks reveals a concerning reality: 10% of the average company’s GenAI apps are considered high-risk.
What if the very AI tools driving workplace productivity were also opening the backdoor to cyber threats? That’s the concern raised by Palo Alto Networks in its latest State of Generative AI 2025 report. The cybersecurity giant reveals that on average, 10% of the 66 GenAI apps used in a typical enterprise today fall under the high-risk category.
Based on data from over 7,000 global enterprises, the report underscores a critical contradiction: as Generative AI (GenAI) adoption accelerates, security controls have not kept pace—particularly across Asia-Pacific and India.
Rapid Growth, Rising Risk
The findings show an 890% spike in GenAI traffic in 2024, with India emerging as a major adopter. Popular apps like Grammarly, Microsoft Power Apps, and Copilot dominate usage across Indian enterprises. But beneath the surface lies growing concern over Shadow AI—unauthorized or unmonitored AI use—making it hard for IT teams to track data exposure and enforce compliance.
“AI is transforming how governments and businesses work,” says Tom Scully, Director at Palo Alto Networks, Asia Pacific & Japan. “But that innovation must be matched with strong oversight. Otherwise, the very tools we celebrate could compromise national security, public trust, and business integrity.”
GenAI Use Cases… and Misuse
- The launch of DeepSeek-R1 in January 2025 triggered a 1,800% spike in related traffic within two months.
- Data loss prevention (DLP) incidents tied to GenAI more than doubled in 2025, now accounting for 14% of all security incidents.
- AI models remain vulnerable to jailbreak attacks, producing harmful or unsafe outputs.
Securing the Future of AI
Palo Alto Networks recommends three key actions for businesses:
- Establish visibility and control of GenAI use across the organisation.
- Safeguard sensitive data with real-time monitoring and policy enforcement.
- Adopt Zero Trust architectures to defend against evolving AI-powered attacks.