Nandini Roy Choudhury, writer
Brief news
- Tech companies are investing in “sovereign AI” to enhance competitiveness and comply with local data sovereignty regulations, emphasizing local infrastructure for data storage.
- European lawmakers are concerned about reliance on U.S. technology, prompting initiatives to develop AI systems reflective of local cultures and languages.
- Local IT firms are collaborating to create regional AI solutions, driven by regulations like GDPR, to ensure data privacy and cultural relevance.
Detailed news
Portugal — Tech companies are putting more money into creating AI models that are “sovereign” in nature. This is because they want to be more competitive by putting more emphasis on local infrastructure.
The concept of “data sovereignty” states that individuals’ personal information should only be kept on servers located within their home country or continent.
“Sovereign AI is a relatively new term that popped up in the last year or so,” Chris Gow, the EU public policy head for IT networking giant Cisco located in Brussels, told CNBC.
Many of the largest LLMs in use today, such as Claude from Anthropic and ChatGPT from OpenAI, keep their data in U.S. data centers and handle queries through the cloud.
European lawmakers and regulators are worried that their continent’s competitiveness and, more worrisomely, technical resilience might suffer as a result of its reliance on American technology.
What is the origin of the term “AI sovereignty”?
Historically, Europe has prioritized the idea of data and technology sovereignty. It was partially brought about by companies responding to new rules.
For instance, businesses are obligated to secure, comply with, and respect the right to privacy of their users in accordance with the EU’s General Data Protection Regulation. Concerns about the security of transferring personal data across borders have been stoked by high-profile incidents in the European Union.
In 2020, the European Court of Justice struck down a data-sharing arrangement between the European Union and the United States, citing concerns that it failed to provide adequate protection comparable to that of the General Data Protection Regulation (GDPR) within the EU. To guarantee the secure transfer of data between the European Union and the United States, the two regions’ data privacy frameworks were established last year.
The localization of cloud infrastructure, which stores and processes data for several internet services, has been driven by these political developments.
French cloud provider OVHCloud’s worldwide president of marketing and operations, Filippo Sanesi, claimed that customers “understand the value of having their data in Europe, which are subject to European legislation.” This has led to high demand for the company’s European-located infrastructure.
“As the idea of data sovereignty develops and its meaning is better understood, an increasing number of companies are realizing the value of storing their data locally, under a specific jurisdiction and set of rules,” Sanesi told CNBC. He also mentioned that they have a large amount of data. “In certain nations, this information is considered private and subject to their laws.”
“Now that you have this data, you can start making AI products and services. Those services should be autonomous, meaning they should be run, deployed, and developed by local talent for the benefit of the local workforce or companies.”
As far as Cisco’s Gow is concerned, regulators have not yet propelled the AI sovereignty movement. Instead, it has originated from private firms, he added, who are expanding their data center operations in Europe. These centers house massive amounts of computer equipment that power cloud-based AI applications.
The “industry naming it that, than it is from the policymakers’ side,” according to Gow, is the primary force behind sovereign AI. “The regulator side has not yet used the ‘AI sovereignty’ terminology.”
A lot of countries are trying to establish national AI policies because they see AI as “the future” and a “massively strategic technology,” as Gow put it.
Governments throughout the world are putting a lot of effort into supporting local IT companies, ecosystems, and the crucial backend infrastructure needed for AI services.
The bandwidth required for an AI job is twenty times more than that of a conventional workload, indicated Gow. Companies can’t succeed without competent employees, therefore empowering the workforce is equally important, says Gow.
But the data is the most essential thing. “In language, what you’re seeing is a lot of effort from that side to consider training LLMs,” Gow explained.
“Showcasing principles”
This summer, Italia 9B became the first LLM in Italy to be trained on data pertaining only to the Italian language.
In order to make sure that the AI systems in Italy are more anchored in the local languages, culture, and history, the project’s goal is to keep results in a specific jurisdiction and use data from residents in that area.
A country’s values and language, as well as those of an enterprise, are what “sovereign AI is all about reflecting,” The leader of enterprise sales for the Eastern and Central European region at Nvidia, David Hogan
When asked about the current state of border models, Hogan emphasized that the main problem is that they have been trained using data from the West.
For instance, according to Hogan, authorities in Denmark—a country with a significant presence of Nvidia—are worried that AI systems providing essential services like healthcare and telecommunications won’t be “reflective” of Danish culture and values.
Denmark released a historic white paper on Wednesday explaining how businesses can use artificial intelligence in a way that complies with the soon-to-be-enacted European Union AI Act, the first such law in the world. Other EU members are expected to follow and accept the paper as a framework.
There is likely less than 2% of the data trained on your language, let alone your culture, if you happen to be in a European country that isn’t one of the major language countries spoken internationally, according to Hogan.
A paradigm change and how regulation drove it
That doesn’t mean rules haven’t played a role in encouraging tech companies to consider constructing regional AI infrastructure in Europe.
Sanesi of OVHCloud noted that rules such as the General Data Protection Regulation (GDPR) in the European Union sparked a lot of interest in moving data processing to a specific location.
Local European IT companies are also supporting the idea of AI sovereignty.
Ecosia, located in Berlin, and Qwant, based in Paris, have launched a partnership to build a European search index from the ground up. The goal is to provide better results in French and German.
At the same time, Orange, a French telecom company,claims to be in talks with multiple AI model foundations to develop a “sovereign AI” model for its customers’ smartphones that better represents their native language and culture.
Making our own LLMs is a complete waste of time. Therefore, the question of “how can we work with current providers to make it more local and safer?” is currently receiving a lot of attention. Speaking to CNBC, Bruno Zerbib—the chief technical officer at Orange—was quoted.
The utilization of local processing on a phone rather than cloud processing for AI data is feasible in many circumstances, according to Zerbib. With these aspirations for a sovereign AI model, Orange has not yet decided on a partner.
Source: CNBC news
