Palantir chief executive Alex Karp on Wednesday said many companies are paying for
artificial intelligence (AI) services that generate little value while exposing their most valuable business information.
"Every single enterprise in this country, these people are livid. They are paying for tokens that create no value. These people are stealing the weights and alpha of my business," Karp said during an interview with CNBC.
His blunt remark comes at a time when businesses are reassessing both the cost and strategic implications of AI adoption. As enterprises seek alternatives to expensive frontier models from
OpenAI and Anthropic, a new generation of lower-cost Chinese models is gaining traction. At the same time, advances by Chinese AI firms in specialised areas such as cybersecurity are raising fresh questions about data security, intellectual property and technological leadership.
Cheap AI is reshaping enterprise choices
The growing appeal of Chinese models comes at a time when enterprises are reassessing the economics of AI adoption.
According to a Reuters report, soaring AI bills are pushing companies to prioritise affordability over marginal improvements in model performance. The shift has prompted a broader pricing battle across the industry, with cheaper alternatives increasingly competing with premium frontier models.
Supporters of the open-source approach argue that cost advantages could accelerate adoption.
Speaking to Rest of World, Tiezhen Wang, former head of the Asia-Pacific at Hugging Face, said many companies initially build products using proprietary models before moving to open-source alternatives as they scale. He added that switching later could allow firms to reduce spending on AI tokens substantially.
Why stronger cyber capabilities raise new questions
The concern, however, is no longer limited to price. Chinese models are also beginning to narrow the performance gap with leading US systems in specialised areas, including cybersecurity.
China's Z.ai (formerly Zhipu AI) in its latest open-weight model, GLM-5.2, demonstrated bug-finding and vulnerability-detection capabilities approaching Anthropic's Mythos-class systems.
Security researchers cited by the Wall Street Journal in a recent report said the model can match Mythos in some software vulnerability benchmarks, even though it continues to trail leading models from Anthropic and OpenAI in broader reasoning and general-purpose tasks.
The findings suggest that Chinese AI developers are making rapid gains in one of the most strategically significant areas of frontier AI development: identifying software vulnerabilities before they can be exploited by attackers.
This development matters because cybersecurity is increasingly viewed as one of the most strategically important areas of frontier AI development. Models that can help security teams discover vulnerabilities can also raise concerns about how such capabilities might be used if deployed for offensive cyber activities.
The Mythos episode and Washington's AI dilemma
The growing capabilities of Chinese AI models come as Washington sharpens its focus on maintaining technological leadership. The US government's AI strategy is built around three broad priorities: accelerating innovation, expanding AI infrastructure and strengthening national security safeguards around advanced technologies.
The White House's America's AI Action Plan describes AI leadership as central to economic competitiveness and national security, while calling for measures to prevent advanced technologies from being misused or stolen by adversaries.
The policy debate became more visible earlier this year when the US government restricted access to Anthropic's advanced Mythos 5 and Fable 5 models over national security concerns and export-control requirements.
Although some restrictions have since been eased, the episode did showcase a growing challenge for policymakers. While the United States wants its AI companies to remain globally dominant, it is also seeking to control access to technologies that could have strategic or security implications.
That tension has become more pronounced as Chinese firms expand their open-source strategy.
Data, trust and the cost of AI adoption
Open-source models allow developers to download, modify and deploy systems without relying on a single provider. While the approach may encourage innovation at lowers costs, it can make oversight more difficult.
Karp's remarks tapped into a broader concern among businesses about how AI systems handle proprietary information. As enterprises feed customer records, internal documents and operational data into AI tools, questions are increasingly being raised about intellectual property, data governance and competitive advantage. "If it was so valuable, let's say I can make you $1 billion tomorrow. Wouldn't I say I'll make you $1 billion and I want 30 percent? Why are they charging for tokens if it's so valuable?" he asked.
Those concerns are becoming harder to ignore as cheaper and more capable alternatives emerge.
 "Why businesses are choosing cheap Chinese AI models over AI giants")
