Companies Turn to Sovereign AI and Open Models Amid Data Privacy Concerns with OpenAI and Anthropic

Data privacy has become a critical factor in AI vendor selection, driven by fears that frontier providers like OpenAI and Anthropic may use customer intellectual property to improve their models or build competing products. A significant point of contention arose in June when Anthropic mandated a 30-day retention policy for chats involving its most powerful models, Fable and Mythos, triggering strong negative reactions from corporate clients.

Sovereign AI Adoption

In response to these concerns, companies are diversifying their AI usage to reduce dependency on leading providers. This shift is encapsulated by the term sovereign AI, which refers to a company’s ability to control its entire AI stack. This includes:

• Owning the chips used for AI workloads.
• Setting up proprietary cloud infrastructure.
• Using open-source models that can be downloaded and run in private environments.

While historically associated with government initiatives, sovereign AI has increasingly entered the corporate lexicon this year as businesses seek greater autonomy over their data and AI operations.