Overview
- Palantir CEO Alex Karp renewed sharp public criticism on Monday, saying hosted frontier models are “trying to drug addict us” and accusing some labs of seeking to capture the means of production.
- Karp paired his attacks with a shareholder letter that described parts of the industry as having “Marxist overtones,” and he used Palantir’s strong Q2 results—$1.9 billion in revenue and $1.1 billion in profit—to push an AI sovereignty argument.
- Frontier labs including OpenAI and Anthropic have publicly denied using customer data to train models and say businesses must opt in for model improvement, a claim Karp says is insufficient to protect enterprise IP.
- Companies, cloud partners and governments are responding by exploring open-weight distributions, on-premise deployments, tighter contracts and AI FinOps teams to retain control of models, data and compute.
- The dispute builds on recent pressures such as Apple’s lawsuit over employee moves to OpenAI and the rise of lower-cost Chinese models, and it could reshape procurement, vendor economics and national-security policy around AI.