If You’re an Educator—Avoid Using the DeepSeek AI Tool

The free AI tool DeepSeek has recently gained a lot of attention, but it’s important to understand the risks before jumping on the bandwagon. This post provides a quick overview of DeepSeek’s capabilities and outlines why it’s not a safe or appropriate choice for school environments.

What is DeepSeek AI?

A Chinese AI company called DeepSeek has launched an AI model that is just as good as most of the leading US AI companies, but at 10 percent of the cost. Its release shook the tech world, with AI stocks losing a trillion dollars in market value. Here’s a breakdown of why this happened:

1. Training Cost Efficiency. We don’t have specific numbers on how much R1 or o-1 cost to train. The 5.6 million dollar figure being reported is misleading because that’s the cost of the final training run for Deepseek’s base model. DeepSeek also used a different training method that doesn’t require as much data collection.

2. Reinforcement Learning. Unlike many U.S. models, which rely on supervised fine-tuning, DeepSeek uses reinforcement learning. This approach allows the model to learn and improve independently, reducing training costs and resources.

3. Open-Source Accessibility. DeepSeek’s R1 model is open-sourced and free to use, unlike OpenAI’s models, which require paid licenses.

4. Cost-Effective Hardware. DeepSeek relies on older, cheaper NVIDIA chips rather than the high-performance, expensive chips used by OpenAI. This raises questions about whether U.S. companies are overspending or inefficient in their budgeting.

Why You Should Avoid DeepSeek—Major FERPA Violations

While DeepSeek’s release has showcased impressive technology and affordability, it comes with significant risks that make it unsuitable for educators or anyone handling sensitive information:

1. User Data Retention. DeepSeek explicitly states that it retains user data even after accounts are deleted.

2. Monitoring Inputs and Outputs. The app reserves the right to monitor all user inputs and outputs, including sensitive and strategic information.

3. Unilateral Updates to Terms. DeepSeek has the right to update its terms and conditions without user consent which raises serious privacy concerns.

4. Chinese Law Governance. Since the app operates under Chinese law, the Chinese government could potentially access user data without restriction.

Implications for Education

If you work in education, the risks associated with DeepSeek should be a red flag. The app’s practices clearly violate FERPA regulations, and the retention of user data — even after account deletion — is highly problematic. As a result, we have blocked DeepSeek in our district to ensure compliance with data privacy standards and to protect our students and staff.

Final Thoughts

DeepSeek’s release has undeniably disrupted the AI landscape, but the privacy and data security risks far outweigh its benefits. While it may be tempting to explore this free and open-source tool, for schools and educators, DeepSeek is not a viable option.