Archive · 10 Aug 2026
Palantir CEO: AI Firms Trying to 'Drug Addict' Businesses
Think of two kinds of companies.
One kind, like Palantir, helps businesses use data and computers.
The other kind, like Anthropic and OpenAI, makes very smart computers called AI.
The boss of Palantir, Alex Karp, says those AI makers are trying to make us hooked on their machines, like getting addicted to candy, and that they want a future where we own nothing and businesses stop being profitable.
Some AI companies say, 'Don't worry, your secrets are safe with us,' explaining they do not use customer data to teach their computers.
Mr. Karp does not believe them.
He also likes 'open-weight' AI models, which are like recipes anyone can see and change, and he and big companies like Nvidia and Microsoft wrote a letter saying these open models are very important.
A report from the UK says China's open AI models are getting almost as good as the best American ones.
So companies are arguing about who should control smart computers.
Palantir CEO Alex Karp said in a CNBC interview that frontier AI companies like Anthropic and OpenAI are trying to 'drug addict' us to a future they believe they control.
Karp argues businesses are being encouraged to pour money into AI models and compute without retaining control over their data, intellectual property, or the value created by those systems.
Karp, who mentioned Anthropic CEO Dario Amodei, warned that the industry's direction could leave businesses unprofitable, workers without jobs, and adversaries winning.
This is not the first attack: in July Karp called the AI industry 'effing insane' and slammed the token-based business models of OpenAI and Anthropic; both companies say customer data is secure and not used to train models, with OpenAI requiring explicit opt-in.
Karp backs open-weight AI models — Palantir joined Nvidia and Microsoft in an open letter calling them critical to national security — while a UK study found China's leading open-weight models like Z.ai's GLM-5.2 and DeepSeek V4-Pro are closing the gap with frontier systems.
- Who
- Alex Karp, CEO of Palantir, criticized frontier AI companies Anthropic and OpenAI and referenced Anthropic CEO Dario Amodei; Nvidia, Microsoft, and the UK's AI Security Institute are also involved.
- What
- Karp accused AI firms like Anthropic and OpenAI of trying to 'drug addict' businesses into dependence on them, and reiterated his support for open-weight AI models.
- When
- In a recent CNBC interview; Karp's earlier 'effing insane' criticism came in early July, and Palantir signed the open letter late last month.
- Where
- The interview aired on CNBC; Palantir is a US-based company, and the cited research on Chinese models comes from the UK's AI Security Institute.
- Why
- Karp argues enterprises risk becoming dependent on AI labs while giving up control over their intellectual property and spending heavily on AI systems without a clear return.
Karp and Palantir
Frontier AI labs (Anthropic, OpenAI)
AI business models and dependence
Karp and Palantir
Karp says AI firms like Anthropic and OpenAI are trying to 'drug addict' businesses into paying for AI without controlling their data, intellectual property, or returns.
Frontier AI labs (Anthropic, OpenAI)
Anthropic and OpenAI say customer data is secure and not used to train models; OpenAI requires businesses to explicitly opt in for their data to be used for model improvement.
Open-weight vs. closed frontier models
Karp and Palantir
Karp backs open-weight AI models, which let developers customise and run models independently, and argues with Nvidia, Microsoft and others that they are critical to national security.
Frontier AI labs (Anthropic, OpenAI)
Frontier labs like Anthropic and OpenAI offer closed models through their own services, which Karp says leaves businesses dependent on them and without control over the value created.
- Critic
- Alex Karp, CEO of US-based Palantir
- Targets
- Anthropic (Claude maker) and OpenAI (ChatGPT maker)
- Venue
- CNBC interview
- Previous criticism
- In July, Karp called the AI industry 'effing insane' and slammed token-based business models
- Data security stance
- Anthropic and OpenAI say customer data is secure and not used to train models; OpenAI requires explicit opt-in for data use in model improvement
- Open-weight support
- Palantir joined Nvidia, Microsoft and others in an open letter calling open-weight AI models critical to national security
- Key study
- UK's AI Security Institute found China's Z.ai GLM-5.2 and DeepSeek V4-Pro rival frontier models released 4-7 months earlier; the gap was 6-10 months for most of 2025
- Mentioned executive
- Dario Amodei, CEO of Anthropic
