A recent technological achievement by Chinese AI lab DeepSeek is being hailed as a significant win for the open-source artificial intelligence movement.
Experts suggest that this development strengthens the case for open-source AI models as viable alternatives to proprietary systems developed by companies like OpenAI.
DeepSeek’s release of its R1 open-source reasoning model last month sparked widespread industry discussion. The model claims to rival the performance of OpenAI’s proprietary AI while using a more cost-effective and energy-efficient training process.
Industry experts believe that DeepSeek’s success highlights the growing maturity of open-source AI. Seena Rejal, Chief Commercial Officer of AI startup NetMind, noted that the achievement demonstrates how open-source models can compete with—and sometimes even outperform—closed systems.
“DeepSeek R1 has demonstrated that open-source models can achieve state-of-the-art performance, rivaling proprietary models from OpenAI and others,” Rejal said in an interview with CNBC. “This challenges the belief that only closed-source models can dominate innovation in this space.”
Yann LeCun, Meta’s chief AI scientist, echoed Rejal’s sentiment, emphasizing that DeepSeek’s accomplishment is a victory for open-source AI rather than a geopolitical win for China. LeCun highlighted Meta’s own contributions to open-source AI through its Llama model and pointed to the broader benefits of collaborative, transparent research.
DeepSeek’s accomplishment comes amid a global trend toward adopting open-source AI technologies. In Europe, initiatives such as OpenEuroLLM are bringing together academic institutions, companies, and data centers to develop multilingual AI models. This movement aims to promote “AI sovereignty,” reducing dependency on Silicon Valley firms.
China, too, has embraced open-source AI, partly due to US export restrictions on advanced chips. Many Chinese firms, including DeepSeek, see open-source technologies as a way to accelerate innovation and broaden adoption of AI models.
Despite its benefits, open-source AI carries inherent risks. Experts caution that the openness of these systems makes them vulnerable to cyber exploitation, including prompt jailbreaking, software supply chain attacks, and data leaks.
Cisco’s AI safety research team recently identified critical vulnerabilities in DeepSeek’s R1 model, warning that it was susceptible to harmful prompts with a 100% attack success rate during tests.
“DeepSeek R1 was purportedly trained with a fraction of the budgets that other frontier model providers spend. However, it comes at a different cost: safety and security,” Cisco researchers Paul Kassianik and Amin Karbasi wrote in their report.
Data privacy is another major concern. DeepSeek’s R1 processes data through its website and app, with information routed directly to China. This has raised fears of potential surveillance by Chinese authorities.
DeepSeek’s advancement has had a ripple effect on the broader AI market. The announcement led to a sharp sell-off in Nvidia’s stock, with concerns that reduced demand for high-performance computing infrastructure could hurt chipmakers.
However, analysts have largely dismissed the fears, suggesting that DeepSeek’s breakthrough could instead expand AI applications and drive demand for chips. Nvidia spokespersons praised the development, noting that it demonstrates how new AI models can be created using compliant technology and available hardware.
CNBC and The Motley Fool contributed to this report.
