What has happened?

The Chinese tech company DeepSeek recently launched a new AI model that –unexpectedly for many – can compete with industry leaders like OpenAI and Google, but at a lower cost and with partially open-source technology. This could accelerate global AI adoption and change who can develop and use advanced AI solutions, as well as how they are applied.

However, the launch has also raised concerns about data security and transparency. Like other major AI models, DeepSeek processes and utilizes input data, but since the company operates under Chinese jurisdiction, questions remain about oversight and control. Despite DeepSeek’s announcement of European data centers that manage information locally and in isolation, skepticism persists.

This is shaking up the playing field among hyperscalers and major tech companies – how will they respond? Despite an ongoing investigation into alleged data theft from OpenAI, Microsoft has added DeepSeek to Azure—a clear sign that they prioritize having the best AI models on their platform, regardless of legal questions in the background. Meta may also be forced to adapt, especially after Mark Zuckerberg recently emphasized that the company plans to “invest heavily” in AI and open-source. With DeepSeek now in the game, competition is intensifying, which could push Meta to open up even further and accelerate its AI initiatives.

Since its launch, OpenAI and Microsoft have continued to distance themselves from DeepSeek and other Chinese AI players, while the U.S. has taken measures to restrict Chinese AI investments and data collaborations. This could have long-term consequences for the global integration of AI platforms.

Sofie Perslow offers insights into how Swedish companies should approach these opportunities and risks – and how they may impact corporate strategies, cybersecurity, and the global AI power balance.

1. The AI landscape is shifting – and DeepSeek is part of it

DeepSeek’s new model is more than just a chatbot—it is an AI model that already rivals OpenAI and Google, but at a lower cost and with broader accessibility.

“This development shows that AI dominance is no longer limited to a few Western companies. For businesses and organizations, this means more options but also new challenges to consider,” says Sofie Perslow.

Perhaps most importantly, real-time data analysis and large-scale modeling are now within reach. Previously, AI models had limited capacity to process and draw conclusions from massive data streams in real time. DeepSeek changes that. Industries such as finance, logistics, and cybersecurity – where milliseconds can determine success or failure – can now leverage AI that not only analyzes data but also predicts events before they happen.

2. Data sharing and transparency – a critical issue

Every new technology brings opportunities but also risks, and DeepSeek adds layers of complexity to the risk landscape. As with all AI platforms, understanding data handling practices is crucial. Regardless of the provider, users should assume that input data can be analyzed, utilized, and, in some cases, stored. DeepSeek is no exception. The key question is where the information goes and who has access to it.

“In Sweden, we need to be strategic in our AI adoption. Banning Chinese tech companies isn’t a realistic solution, and doing so could hinder our own development. What is clear, however, is that the public sector and Swedish businesses must take responsibility for understanding the data they share and how it is used. AI models are powerful tools – but they are only as secure as the people using them,” says Sofie Perslow.

It has also become clear that DeepSeek is not as transparent as many initially believed. Training datasets and details about the training data have not been disclosed, meaning the model still operates with “black boxes”. While developers can use and fine-tune it, understanding exactly what drives its behavior remains challenging.

In response to growing criticism, DeepSeek has started communicating more about its security measures and data protection. However, it remains unclear whether they will open up their training data in the same way as some other AI players.

However, the Hugging Face community is working to address this issue. They are engaging developers to analyze and document how the model functions, aiming to make it more transparent and fully open. Through crowdsourcing and community-driven initiatives, they are attempting to reconstruct parts of the training data. If successful, DeepSeek could become a truly open alternative.

3. AI bias and censorship – different models, different priorities

All AI models have built-in filters and constraints, whether due to technical limitations or deliberate choices by developers. DeepSeek may have different priorities than OpenAI or Google in terms of what information is emphasized or downplayed. This affects everything from search results to decision-making support – something businesses must be aware of when selecting AI platforms.

4. AI is a tool – not a magic solution

DeepSeek has an advanced model, but that doesn’t mean it is always correct. AI models can misinterpret, hallucinate, and confidently produce inaccurate information. It is essential to combine AI adoption with critical thinking and fact-based analysis.

5. Sweden needs a future-proof AI strategy

AI development will not slow down—on the contrary, it is accelerating. We are witnessing a global AI arms race where new players emerge unexpectedly.

“The question is not whether AI will change the world, but how Sweden and Europe can ensure we are prepared. This requires a combination of technical expertise, rapid implementation, ethical considerations, and strategic decisions on data sharing and AI usage,” says Sofie Perslow.

How is HiQ approaching DeepSeek?

“We approach DeepSeek the same way we do all AI platforms – with curiosity, but also with a strong sense of realism. The technology is impressive, but that doesn’t mean it can be used indiscriminately. Security and business value always come first. Our approach is simple: we test, analyze, and make decisions based on facts and data – not hype. DeepSeek is an exciting player, but AI models are tools, not magic solutions.”

“DeepSeek is a reminder that the AI landscape is evolving rapidly. The only certainty about AI is that development will never be slower than it is today. For Swedish businesses and public organizations, the question is not whether to use AI – but how to choose the right AI solutions and implement them responsibly, with the right insights and strategy,” concludes Sofie Perslow.

Do you have questions about AI strategies and implementation? Contact us!