Many organizations are grappling with how to get the most value from AI without risking data security or violating laws. The push for affordable, high-performance AI models has sparked debates about the true costs and hidden risks involved. Companies need to rethink their approach to AI vendors, especially when it comes to data privacy and geopolitical concerns.

The Shift Toward Cost-Effective AI Models

For over a year, the AI industry focused on building bigger models with more parameters, equating size with success. Success was often measured by parameter counts and benchmark scores, which don’t always translate to practical business value. Recently, there’s been a shift as some companies explore smaller, cheaper AI models that can deliver good results without the hefty price tag.

DeepSeek, a China-based AI lab, has gained attention for demonstrating that large language models don’t have to be expensive to be effective. Their approach challenged the idea that only billion-dollar investments can produce top-tier AI. This message resonated with businesses eager to reduce costs in AI pilots and deployments.

Bill Conner, CEO of Jitterbit and former advisor to security agencies, notes that reports of DeepSeek’s low training costs have reignited conversations around efficiency and “good enough” AI. But these cost savings come with their own set of complex issues—particularly around data security and sovereignty.

Data Sovereignty and Geopolitical Risks

The excitement over affordable AI models collided with geopolitical realities when disclosures revealed DeepSeek’s data practices. It turns out the company was storing data in China and sharing it with Chinese intelligence agencies. For Western companies, this raises serious concerns about data security and legal compliance.

It’s not just about following privacy laws like GDPR or CCPA anymore. The bigger issue is national security. When sensitive data is stored or shared across borders, it can be accessed by foreign governments or malicious actors. This creates risks that can’t be mitigated by simple privacy policies.

Integrating AI models often involves connecting them to proprietary data systems, customer info, or intellectual property. If the AI provider has a back door or shares data with foreign governments, a business’s sovereignty is compromised. This could lead to breaches, legal penalties, or even sanctions violations.

Conner warns that DeepSeek’s links to military procurement and potential export control breaches should serve as warnings for company leaders. Using such technology might inadvertently entangle a business in international sanctions or supply chain issues, risking more than just financial loss.

Today, choosing an AI vendor is about more than code performance. It’s about trusting their legal and ethical standards. Companies need to carefully evaluate the risks involved in using AI solutions from different jurisdictions, especially when data security and national interests are at stake.

As AI continues to evolve, organizations must strike a balance between cost efficiency and maintaining control over their data and security. Making informed choices now can help prevent costly problems down the line, ensuring AI provides value without exposing the business to unnecessary risks.

Inspired by

• https://www.artificialintelligence-news.com/news/balancing-ai-cost-efficiency-with-data-sovereignty/

Sources

• interpol.int
• gov.uk
• jitterbit.com
• developer-tech.com