Recent research indicates that open-weight AI models, such as DeepSeek, are nearing the hacking capabilities of advanced closed models like OpenAI's Daybreak. In simulated tests, DeepSeek completed 73% of tasks, demonstrating significant potential for both attackers and defenders in cybersecurity. The cost-effectiveness of open models further enhances their appeal, with much lower operational costs compared to proprietary options.
The findings underscore a shift in the cybersecurity landscape, as AI continues to automate hacking tasks and reveal vulnerabilities. While the implications of this trend are still debated, the ability of open models to match closed ones raises concerns about how criminal hackers might leverage these technologies, necessitating a reevaluation of security strategies by companies and developers.