Palantir warns companies to keep data in-house, avoid over-relying on LLM providers

Palantir has released a nine-point manifesto urging corporations to maintain control over their proprietary data rather than handing it to commercial large language model providers. The document centers on the concept of 'AI sovereignty,' stressing that ownership of data and model weights is critical to a company's long-term competitiveness. Palantir also cautioned against 'tokenmaxxing,' a term it uses to describe excessive AI spending without a clear or purposeful strategy. The company advised that AI-related decisions should be driven by sound judgment rather than politics or favoritism.