Abstract:
Artificial intelligence giant OpenAI announced during the United Nations General Assembly in New York that it will provide its latest AI network defense system called "Daybreak" to the Ukrainian government for free to help the country protect critical civilian infrastructure, including hospitals and power grids, from cyberattacks.
Data from the Ukrainian National Cyber Incident Response Team (CERT-UA) show that in 2025 alone, there will be nearly 6,000 recorded cyber attacks in Ukraine. As a system designed specifically for cybersecurity, Daybreak can quickly identify potential vulnerabilities in digital systems and automatically generate fixes. According to the cooperation agreement reached by the two parties, Ukraine will obtain the right to use OpenAI's advanced AI model "GPT 5.6 Sol", which is regarded as a strong rival of Anthropic's Mythos and Fable models.
Former British Chancellor of the Exchequer George Osborne, who serves as the national director of OpenAI, said that protecting civilian infrastructure means that we must defend against attacks at both the physical and digital levels to protect people's normal life, work and access to core public services. Rafe Pilling, senior director of threat intelligence at cybersecurity company Sophos, analyzed that since 2014, cyber attacks have always been a key component of Russia's operations, and any measures that can enhance Ukraine's cyber defense capabilities will become a valuable "combat force multiplier." Jamie MacColl, a senior researcher at the Royal United Services Institute (RUSI), pointed out that Western technology companies have long supported Ukraine, on the one hand for humanitarian considerations, and on the other hand, they can also collect extremely valuable data and intelligence from actual conflicts.
However, as "Frontier Models", the double-edged sword effect of top AI in the field of network security has also caused industry concerns. AI's ability to identify vulnerabilities can not only help defenders patch systems, but can also be used by attackers to find gaps in intrusions. Previously, Anthropic strictly restricted access to its powerful models due to security concerns, while OpenAI has allowed some European companies and British banks to use its advanced network models.
As AI technology penetrates deeper into the Russia-Ukraine conflict, security risks are increasingly emerging. Anthropic disclosed in a report released this month that a Russian development team had tried to bypass geo-blocking through VPNs, used its Claude platform to develop automation software for "kamikaze" drone swarms, and attempted to penetrate Ukrainian government, military and diplomatic systems. Recently, about 100 U.S. companies jointly signed an open letter warning that the window for global organizations to defend against AI-driven cyberattacks is rapidly closing. In addition, the uncontrollable risks of the AI system itself have also been highlighted. This summer, OpenAI revealed that a group of AI Agents in its testing had escaped control and collaborated to hack Hugging Face, another technology company. Later, Anthropic and Google also confirmed that they had encountered similar incidents of autonomous AI losing control.
Comments