LISBON, Portugal — Tech giants are increasingly investing in the development of so-called “sovereign” artificial intelligence models as they seek to become more competitive by focusing more on local infrastructure.
Data sovereignty means that people’s data should be stored on the infrastructure of the country or continent where they live.
Chris Gow, head of EU public policy at IT networking giant Cisco in Brussels, told CNBC: “Sovereign AI is a relatively new term that has emerged in the last year or so.”
Currently, many of the largest large-scale language models (LLMs), such as OpenAI’s ChatGPT and Anthropic’s Claude, use data centers in the United States to store data and process requests through the cloud.
This has caused concern among European politicians and regulators, who believe that reliance on American technology is detrimental to the continent’s competitiveness and, more worryingly, technological resilience.
Where does “artificial intelligence sovereignty” come from?
concept Data and technology sovereignty This was previously on the European agenda. This is partly the result of businesses reacting to new regulations.
For example, the European Union’s General Data Protection Regulation requires companies to process user data in a secure and compliant manner and respect their privacy rights. High-profile cases in the EU Questions have also been raised about whether European citizens’ data can be safely transferred across borders.
The European Court of Justice invalidated the EU-US Data Sharing Framework in 2020 on the grounds that the agreement did not provide the same level of protection as guaranteed by the General Data Protection Regulation (GDPR) within the EU. last year’s EU-US Data Privacy Framework Established to ensure that data can flow securely between the EU and the United States
These political developments ultimately drove the localization of cloud infrastructure, where data is stored and processed for many online services.
Filippo Sanesi, global head of marketing and operations at OVHCloud, said the French cloud company is seeing a lot of demand for its infrastructure in Europe because they “understand the value of having data in Europe that is subject to European legislation.”
Sanesi told CNBC: “As the concept of data sovereignty becomes more mature and people understand what it means, we are seeing more and more companies recognizing that data should be stored locally and under specific jurisdictions and governance. importance. “We have a lot of data,” he added, “that is sovereign in specific countries. “
“Now, with this data, you can actually make products and services for artificial intelligence, and those services should be sovereign and should be controlled, deployed and developed locally by local talent for local populations or businesses.”
Cisco’s Gao said regulators haven’t pushed for AI sovereignty, at least not yet. Instead, it comes from private companies that are opening more data centers in Europe that contain vast amounts of computing equipment to support cloud-based artificial intelligence tools, he said.
Gao said sovereign AI “is driven more by industry naming than policymakers.” “You haven’t seen regulators use the term ‘AI sovereignty’ yet.”
Gao said countries are promoting the idea of AI sovereignty because they recognize that AI is “the future” and a “large-scale strategic technology.”
Governments are focusing on developing domestic technology companies and ecosystems, as well as the all-important back-end infrastructure that supports AI services.
“Artificial intelligence workloads use 20 times the bandwidth of traditional workloads,” Gow said. Gao said it’s also about supporting the workforce, as companies need skilled workers to succeed.
However, the most important thing is the data. “What you’re seeing is that there are quite a few attempts to consider localized data and language for LLM training,” Gao said.
“Reflect values”
In Italy, The first LL.M. with specialized training in Italian language materialscalled Italia 9B, launched this summer.
The Italian project aims to localize results within a given jurisdiction and rely on data from the region’s citizens so that the results produced by the AI system are more rooted in local language, culture and history.
“Sovereign AI is designed to reflect the values of an organization, or equally the values and language of your country,” said David Hogan, head of enterprise sales for the chip manufacturing giant EMEA. NVIDIAtold CNBC.
Hogan added: “The core challenge is that most cutting-edge models today are primarily trained on Western data.”
Hogan said that in Denmark, for example, where Nvidia has a strong presence, officials are worried that important services such as health care and telecommunications provided by artificial intelligence systems will not “reflect” local Danish culture and values.
On Wednesday, Denmark released a landmark white paper outlining how companies can use artificial intelligence in line with the upcoming EU Artificial Intelligence Act, the world’s first major AI law. The document is intended to serve as a blueprint for other EU countries to follow and adopt.
“If you’re in a European country that’s not one of the major languages spoken internationally, then probably less than 2 percent of the data is trained on your language, let alone your culture,” Hogan said.
How supervision promotes changes in thinking
That’s not to say regulations haven’t proven to be an important factor in prompting tech giants to consider more of building localized AI infrastructure in Europe.
OVHCloud’s Sanesi said regulations such as the EU’s GDPR have fueled interest in outsourcing data processing to specific regions.
The concept of artificial intelligence sovereignty has also been recognized by local European technology companies.
Earlier this week, Berlin-based search engine Ecosia and its Paris-based counterpart Qwant announced a joint venture Developing a European search index from scratchdesigned to provide improved French and German results.
At the same time, French telecom operators orange It said it is in discussions with a number of basic artificial intelligence model companies to build smartphone-based “sovereign artificial intelligence” models for its customers to more accurately reflect their own language and culture.
“It doesn’t make sense to build our own LLM. So there’s a lot of discussion now about, how do we work with existing providers to make it more local and more secure?” Orange chief technology officer Bruno Zerbib told CNBC.
“In many use cases, (artificial intelligence data) can be processed locally (on the phone) rather than in the cloud,” Zerbib added. Orange has yet to select a partner for these sovereign AI model goals.
