Synthetic intelligence (AI) has quickly shifted from buzz to enterprise necessity over the previous 12 months—one thing Zscaler has seen firsthand whereas pioneering AI-powered options and monitoring enterprise AI/ML exercise on the earth’s largest safety cloud. Enterprises are more and more adopting AI instruments to reinforce productiveness, automate workflows, and speed up decision-making. Nonetheless, cybercriminals are leveraging the identical know-how to scale subtle assaults, from hyper-realistic deepfakes to superior phishing schemes.
With AI basically altering each how companies function and the way cybercriminals assault, organizations should preserve a present and complete understanding of the enterprise AI panorama. The just-released ThreatLabz 2025 AI Security Report examines the intersection of enterprise AI utilization and safety, drawing insights from 536.5 billion AI/ML transactions within the Zscaler Zero Belief Trade.
The report reveals how enterprises worldwide and throughout industries are utilizing and managing AI/ML instruments, highlighting each their advantages and safety issues. It examines rising dangers related to AI, from cybercriminals weaponizing AI to the safety implications of current AI developments like DeepSeek, whereas offering finest practices for mitigating these dangers.
5 key findings: AI utilization and risk traits
The ThreatLabz analysis staff analyzed exercise from over 800 identified AI/ML functions between February and December 2024. Listed below are the notable findings:
1. AI/ML utilization surged exponentially: AI/ML transactions within the Zscaler cloud elevated 36x (+3,464.6%) year-over-year, highlighting the explosive progress of enterprise AI adoption. The surge was fueled by ChatGPT, Microsoft Copilot, Grammarly, and different generative AI instruments, which accounted for almost all of AI-related site visitors from identified functions.
Zscaler
Determine 1: High AI functions by transaction quantity
2. Enterprises blocked a big proportion of AI transactions: 59.9% of AI/ML transactions have been blocked, signaling issues over information safety and the uncontrolled use of AI functions. As organizations work to determine AI governance frameworks, many are taking a cautious method, proscribing entry to sure AI functions as they refine insurance policies round information safety.
3. U.S. and India drive essentially the most AI/ML site visitors: The US and India recorded the best quantity of AI/ML transactions within the Zscaler cloud, reflecting robust enterprise adoption and a rising concentrate on AI-driven innovation. Different high contributors embody the UK, Germany, and Japan, every exhibiting completely different ranges of AI/ML exercise.
4. Finance & Insurance coverage and Manufacturing dominate AI adoption: The Finance & Insurance coverage (28.4%) and Manufacturing (21.6%) sectors generated essentially the most AI/ML site visitors. Following them, Know-how, Healthcare, and Authorities are integrating AI at various charges as they navigate the wonderful line between adoption and apprehension.
Zscaler
Determine 2: Industries driving the most important proportions of AI transactions
5. AI-driven cyber dangers are escalating: Risk actors are leveraging AI to reinforce phishing campaigns, automated assaults, and create life like deepfake content material. ThreatLabz researchers demonstrated how DeepSeek could be manipulated to rapidly generate phishing pages that mimic trusted manufacturers. Moreover, ThreatLabz uncovered a malware marketing campaign through which attackers created a faux AI platform to take advantage of curiosity in AI and trick victims into downloading malicious software program.
Securing AI and staving off AI threats with Zscaler
The ThreatLabz 2025 AI Security Report gives detailed steerage for enterprises seeking to securely undertake AI whereas minimizing dangers and stopping AI-powered cyberthreats.
Enterprises should rethink safety methods to account for brand new vulnerabilities, expanded assault surfaces, and AI-fueled cyberattacks. Conventional safety approaches reliant on firewalls and VPNs are woefully inadequate towards the velocity and class of AI-powered threats. Enterprises should undertake a zero belief method, eliminating implicit belief, implementing least-privilege entry, and repeatedly verifying all AI interactions.
Zscaler’s zero trust architecture delivers Zero Belief In all places—securing consumer, workload, and IoT/OT communications—infused with comprehensive AI capabilities. Its AI fashions detect and disrupt superior threats, blocking thousands and thousands of assaults every day to reinforce enterprise safety outcomes and mitigate rising dangers.
The report particulars the way to cease AI-powered threats with Zscaler, together with:
- Zero belief structure: Cut back the assault floor by hiding functions and IP addresses from attackers and implementing least-privilege entry.
- AI-powered cyberthreat safety: Detect and block AI-generated phishing campaigns, adversarial exploits, and AI-driven malware in actual time.
- AI-powered information classification and DLP: Use AI-driven classification to detect and shield delicate information throughout Zscaler’s Knowledge Cloth.
- AI-powered app segmentation: Cease lateral motion inside networks, guaranteeing attackers can not simply escalate privileges or entry important methods.
- AI-powered breach prediction: Preempt potential breach situations utilizing generative AI and multi-dimensional predictive fashions.
- Actual-time AI insights: Make use of predictive and generative AI for actionable insights that improve safety operations and digital efficiency.
- AI visibility: Get in-depth visibility into AI software traits and interactions via interactive dashboards.
Get the report
Download the ThreatLabz 2025 AI Security Report for added data-driven insights and evaluation of AI’s impression on cybersecurity, with professional steerage to assist enterprises securely embrace AI and mitigate its dangers.