Page 160 - Cyber Defense eMagazine RSAC Special Edition 2025

Should Enterprises Build Their Own AI Models Instead?

One alternative to using DeepSeek’s API is for enterprises to develop their own AI models based on
DeepSeek’s techniques. Given its open-weight nature, companies with the expertise to fine-tune AI
models may opt for a self-hosted approach that eliminates risks tied to third-party data storage. By taking
control of the AI development process, organizations can reduce their dependence on external providers
and mitigate concerns related to data privacy and security.

This approach may also help enterprises increase compliance, as it allows businesses to fully control the
AI training and inference process, ensuring that sensitive data remains within their own infrastructure.
Additionally, developing an in-house model enables organizations to tailor AI capabilities to their specific
needs, optimizing performance while maintaining strict security protocols.

Seeking Alternatives to Safehouse Data

AI innovation must not come at the expense of security. DeepSeek’s rise reflects the demand for
affordable, high-performance AI, but enterprises must carefully weigh the risks. The model’s China-based
data storage, undisclosed training data, and potential security vulnerabilities introduce compliance
challenges that businesses cannot ignore.

In 2024 alone, the use of generative AI among enterprises increased by 71%. As the adoption of this
technology accelerates, companies must prioritize cybersecurity strategies that include robust security
frameworks, regulatory alignment, and ethical AI practices to harness the benefits of AI while mitigating
risks.

Businesses that proactively address these risks will be best positioned to leverage AI safely and
effectively in a rapidly evolving technology and regulatory landscape.

About the Author

Alix Melchy is the VP of AI at Jumio, where he leads teams of machine
learning engineers across the globe with a focus on computer vision,
natural language processing and statistical modeling. An experienced
AI leader, Melchy has a passion for turning AI-innovation into
enterprise-grade AI systems, fostering the responsible practice of AI
and shaping a secure digital landscape.

Alix can be reached on LinkedIn and at his company’s website
https://www.jumio.com/

160