I'm LongbridgeAI, I can summarize articles.Major firms like Palantir, Nvidia, and Booz Allen Hamilton are reducing reliance on OpenAI and Anthropic models due to data privacy and IP concerns. Tensions rose after Anthropic's policy changes and OpenAI's data usage claims. Companies now demand zero-data-retention guarantees, with some shifting to Microsoft's private AI setups or in-house models. This shift highlights growing corporate caution regarding sensitive data exposure in enterprise AI applications.
Major companies that handle sensitive work, including Palantir (PLTR), Nvidia (NVDA), and Booz Allen Hamilton (BAH), are pulling back from using the most advanced models from Anthropic and OpenAI, The Information reported. The move comes due to rising concerns about how the AI labs handle customer data and intellectual property.
Tension grew after Anthropic changed its policy in June for its main Fable model, giving itself the right to keep customer data for 30 days to help stop misuse. At the same time, OpenAI is facing claims that it may have used a customer's data to help reach a major math breakthrough.
Both OpenAI and Anthropic say they do not train their models on enterprise inputs or outputs unless customers choose to opt in. However, they do collect metadata. OpenAI uses de‑identified usage data to improve its models, while Anthropic uses anonymized, pooled usage data to improve product features but says it does not use it for training.
Growing Corporate Caution
Even as OpenAI and Anthropic grow fast, more companies are drawing lines around what data they will share:
- Palantir is urging clients not to put sensitive IP directly into Anthropic or OpenAI and is demanding zero‑data‑retention (ZDR) guarantees.
- Nvidia uses Anthropic's Fable only for low‑risk tasks and relies on its own Nemotron models for sensitive work.
- Booz Allen blocks employees from using Fable for cybersecurity projects over fears the model could "learn" from proprietary code.
- A major U.S. utility dropped its plans to test Fable for core power grid operations after Anthropic refused to offer nonrevocable ZDR terms.
- Novo Nordisk (NVO) and Northrop Grumman (NOC) use Anthropic or OpenAI only for public or low‑risk tasks, keeping proprietary work on air‑gapped servers with open‑source or in‑house models.
Why This Move Matters
OpenAI and Anthropic offer some of the strongest models for automating office work, which makes them hard for companies to avoid. But both firms are also building their own apps for legal, research, and enterprise tasks. This puts them in direct competition with some of their biggest customers.
As companies weigh the benefits of top‑tier AI against the risk of exposing trade secrets, the push for clear data retention guarantees is becoming a key demand.
It is worth noting that Microsoft (MSFT) is using the moment to pitch its own AI systems to companies worried about privacy. According to the report, Microsoft has offered large clients private, isolated AI setups that run on customer‑managed servers and send no data to outside providers. Some firms are still weighing this multi‑million‑dollar option.
Which Stock Is a Better Buy?
According to TipRanks' Stock Comparison Tool, NVDA and MSFT have Strong Buy consensus ratings, PLTR has Moderate Buy, while BAH has a Hold. Among these stocks, analysts see the most upside in Nvidia at 54.35%, followed by PLTR and MSFT at 20.1% and 14.55%, respectively. Analysts see 9.32% upside in BAH stock.