AI Security: Take the AI to the Data, not Data to the AI

With AI incidents surging by over 56% in 2024, New Zealand businesses face mounting pressure to harness artificial intelligence while protecting their most valuable asset—their data. The solution isn’t avoiding AI security risks altogether, but implementing a smarter approach: bringing AI processing directly to your data rather than sending sensitive information to external AI services.

AI Security for Kiwi Businesses: Keep Your Data Safe While Embracing Innovation

AI security for New Zealand businesses means implementing smart strategies that protect your data while unlocking AI’s potential. This approach focuses on processing AI models within your existing security infrastructure rather than exposing sensitive information to external platforms.

Table of Contents

1. Overview / Background
2. Key Points & Benefits
3. Challenges & Misconceptions
4. Practical Tips
5. Future Outlook
6. Summary & Next Steps

Overview / Background

New Zealand businesses are caught between opportunity and risk. According to Stanford’s 2025 AI Index Report, AI-related security incidents jumped to 233 globally in 2024—a staggering 56.4% increase from the previous year. Meanwhile, Kordia’s 2024 Business Cyber Security Report reveals that 28% of large New Zealand firms identify AI-generated cyber-attacks as their top security concern, even though only 6% of current breaches use AI tactics.

The challenge is clear: 78% of organisations now deploy AI in at least one business function, yet fewer than two-thirds implement adequate safeguards. For Kiwi businesses, this gap represents both risk and opportunity—those who get data privacy right will gain competitive advantage while avoiding costly breaches.

Key Points & Benefits

• Reduced Data Exposure: Processing AI within your existing security setup means sensitive customer data never leaves your controlled environment. A law firm, for example, can analyse client documents using AI without sending confidential information to external servers.
• Faster Threat Detection: Organisations using AI-powered security tools within their own infrastructure reduce threat detection and containment times by up to 12%, according to recent cybersecurity studies. This means quicker responses to potential breaches.
• Cost Control: With New Zealand businesses forecast to spend NZD 5.80 per employee on data security in 2025, keeping AI processing local helps maximise this investment by leveraging existing security infrastructure rather than purchasing additional external AI security measures.

Challenges & Misconceptions

Misconception 1: “On-premise AI is too complex for small businesses.” Reality: Modern AI tools are increasingly designed for local deployment with user-friendly interfaces. The fix is starting small with specific use cases like document analysis or customer service chatbots that work within your current systems.

Challenge: “We don’t have the technical expertise.” Many New Zealand IT providers now offer managed AI services that run within your security perimeter. Partner with local experts who understand both AI implementation and Kiwi privacy requirements under the Privacy Act 2020.  Value 32

Misconception 2: “Cloud AI services are always more secure.” While major cloud providers invest heavily in security, your data is still processed on their systems. Local processing means you maintain complete control over access, encryption, and audit trails—crucial when 16% of New Zealand businesses experienced theft of personally identifiable information in 2024.

Practical Tips

1. Start with Data Classification: Identify which data is most sensitive and prioritise keeping AI processing of this information on-premise. Use role-based access controls to ensure only authorised staff can access AI-processed insights.
2. Implement Hybrid Approaches: Use external AI services for non-sensitive tasks like general market analysis, while keeping customer data, financial information, and strategic planning documents processed locally through on-premise AI solutions.
3. Establish AI Governance Frameworks: Create clear policies covering who can use AI tools, what data can be processed, and how to audit AI decisions. Include regular security assessments to test for vulnerabilities like adversarial attacks.
4. Train Your Team: Educate staff about AI-enhanced phishing attacks, which are becoming more sophisticated. Simple awareness training can prevent the 43% of cyber-attacks that still use phishing as their primary method.

Future Outlook

The next two to five years will see AI security become a defining competitive factor for New Zealand businesses. We expect increasing regulation around AI data handling, similar to Europe’s AI Act, making local AI processing not just smart but potentially mandatory for certain industries.

Edge computing and smaller AI models are making on-premise processing more accessible and cost-effective. However, the threat landscape will also evolve—cybercriminals are already using AI to create more convincing phishing attempts and automated attack systems. Businesses that establish strong AI governance frameworks now will be best positioned to adapt as both opportunities and threats develop.

Conclusion & Next Steps

• AI incidents are rising rapidly, but smart data handling can dramatically reduce your risk
• Processing AI within your existing security infrastructure offers better control and compliance
• Start small with specific use cases and build your AI security capability gradually
• Partner with local experts who understand New Zealand privacy requirements 
• Train your team to recognise AI-enhanced cyber threats while embracing AI’s benefits