DeepSeek Has Shaken Assumptions in the Generative AI Ecosystem
I interviewed international law scholar, Dr. Gianclaudio Malgieri, for a take on DeepSeek, that has become an overnight *sensation* in the West, driving Nvidia stock prices down and creating panic in Silicon Valley.
He is Associate Professor of Law at Leiden University, co-Director of the Brussels Privacy Hub, and Managing Editor of the Computer, Law and Security Review.
What about the use of OpenAI products by US government agencies and potential security risks?
We discussed privacy concerns related to the use of AI tools like DeepSeek and OpenAI's ChatGPT, and the entry into force - in its first requirements and applicability - of the EU Artificial Intelligence Act. Not familiar with the EU AI Act? Here is a primer.
Dr. Malgieri was in a meeting with EU regulators this past Tuesday, including the European Data Protection Supervisor and the Council of Europe, to discuss a range of privacy law issues, but they are not considering any reforms to the GDPR.
He also noted that the Digital Fairness Act, which aims to reduce risks of dark patterns and unfair practices by digital platforms, is still under discussion.
Dr. Malgieri spoke about the challenges of compatibility between large language models and data protection rules, as well as the issue of DeepSeek's being based in China, which raises questions about data transfer outside the European Union.
We spoke about the ongoing investigation by the Italian data protection authorities on DeepSeek, and the recent sell-off of Nvidia stocks, which led to a significant loss in the Nasdaq.
Global investors dumped tech stocks on Monday as they worried that the emergence of a low-cost Chinese artificial intelligence model would threaten the dominance of AI leaders like Nvidia (NVDA.O), opens new tab, evaporating $593 billion of the chipmaker's market value, a record one-day loss for any company on Wall Street. Source: Reuters, read full article here.
AI in Healthcare
We also touched upon the use of AI in various settings, including healthcare, where AI is being touted as a promising tool for faster and more efficient diagnosis and treatment. There is a need for regulators and policymakers to verify the accuracy of these claims.
AI Act and Digital Regulation
Understanding what is prohibited and the need for measures for high-risk AI systems is at the core of the EU AI Act. We also spoke about Mario Draghi's report on competitiveness and its influence on digital regulation and governance.
Malgieri says that an opportunistic narrative of deregulation, favoring industry, was created around Draghi’s report, and the narrative that Europe’s competitiveness is in opposition to the protection of fundamental rights. It is not either/or, in his view.
We also touched upon:
• European Data Protection Board to provide clearer guidelines on the compatibility of large language models with data protection rules and principles.
• Italian Data Protection Authority to continue investigation on DeepSeek's compliance with EU data protection regulations.
• European AI Office to check DeepSeek's respect for fundamental rights.
• Companies and public sector entities to implement and interpret the Artificial Intelligence Act as it becomes applicable.
• Artificial Intelligence Office to provide final details on prohibited AI uses and exceptions under the AI Act.
• High-risk AI system operators (including hospitals, critical infrastructures, border control, schools, banks, and insurance companies) to implement required measures under the AI Act.
• European Union to focus on creating synergies within EU-based AI companies while maintaining clear regulations.
• European Commission to apply and enforce existing fundamental rights protection laws in the digital sphere.
Listen to the interview. Recorded on the 29th January, 2025.
You can buy his book here.
Panic in Silicon Valley? French tech founders and investors discuss:
We are reading:
“Unless you’ve been living under a rock this week, you’ll have probably heard about DeepSeek, the Chinese AI company that just detonated a bomb under U.S. tech stocks.”
Why AI Safety Researchers are Worried About DeepSeek, writes Billy Perrigo.
Yet. There is optimism. Bloomberg.
CNBC on the wild market swing.
DW News on how DeepSeek cost a fraction of the investments in OpenAI.
Semafor takes the safe road.
Markets are spooked, reports Channel4. Gary Marcus reminds us that DeepSeek is not AGI. Victor Gao, a former Chinese government official who is now Vice-President of the Centre for China and Globalisation in Beijing, stresses that a revolution is underway, and none has a monopoly on AI.
“When you buy a company’s stock, you buy its future earnings. It’s still amazing how much can be taken on trust.” Read more
Elsewhere…
User Mag by Taylor Lorenz quotes Hollywood Reporter: “Jesse Eisenberg once tried to drive to Facebook HQ to meet Mark Zuckerberg IRL before being told that “for legal reasons” he could not go there.”
Washington Post
