Ethical Hacking News
Chinese cybersecurity company Qihoo 360 claims it's built an AI bug finder that surpasses Anthropic's Mythos model. The company's multi-agent swarm approach is being touted as a more effective strategy for identifying vulnerabilities, and its tools have already demonstrated impressive capabilities in discovering dormant flaws in software.
Qihoo 360 claims to have developed an AI bug-finder that surpasses Anthropic's Mythos model.The company's approach uses a multi-agent swarm strategy, rather than individualized analysis, to collaborate on threat modeling and attack surface filtering.The multi-agent swarm approach automates building sandbox environments, generating exploit code, and conducting real-world testing to verify vulnerabilities.Qihoo 360 has demonstrated its capabilities by discovering previously dormant vulnerabilities in open-source software and commercial applications.The company is positioning itself as a significant player in the global cybersecurity landscape with its innovative approach to bug-finding and security modeling.
Chinese cybersecurity company Qihoo 360 has made a bold claim, asserting that it has developed an AI bug-finder that surpasses the capabilities of Anthropic's highly touted Mythos model. This assertion was made by CEO Zhou Hongyi during a speech at the 14th Beijing Cybersecurity Conference, which is organized by Qihoo 360 itself.
According to Zhou, Mythos follows a typical large-scale model approach, utilizing immense computational power and robust hardware components to achieve its objectives. Nevertheless, this approach has an inherent drawback - it necessitates the possession of sophisticated model capabilities, which domestically developed models often struggle to match. Zhou posited that China's own models lag behind by 20-30 percent in terms of underlying capabilities.
In light of these challenges, Zhou proposed a novel strategy for Qihoo 360 - utilizing its extensive experience in combating cyber threats and vast malware library to develop security-specific models and agents. By distilling these resources into what he described as a "multi-agent swarm," the company has successfully created an alternative approach that rivals Mythos.
This multi-agent swarm operates on the principle of collaboration, rather than individualized analysis. When confronted with a target, it doesn't focus solely on pinpointing vulnerabilities but instead models threats and filters high-risk attack surfaces before following data flows across files to uncover potential weaknesses. This collaborative strategy is bolstered by the agents' capacity to automatically build sandbox environments, generate exploit code, and conduct real-world testing.
The result of this multi-agent swarm approach is that every identified vulnerability is verified rather than merely suspected, and it also carries out post-task summaries and performance reviews, thereby becoming more intelligent with each use. This level of sophistication surpasses the capabilities of single large models like Mythos.
Qihoo 360 has showcased its capabilities by discovering previously dormant vulnerabilities in open-source software and commercial applications. One notable example is a Windows kernel privilege escalation vulnerability that remained undiscovered for five years and had been recognized by Microsoft. The company's Yitianzhen tool, another AI-powered security solution, automatically simulates potential attacks against an organization's cyber-defenses, suggesting and/or implementing remediations.
Furthermore, Qihoo 360 has established partnerships with local cybersecurity companies to utilize its tools and create a bulwark against the threats posed by Anthropic's controlled conditions. Despite being sanctioned by US authorities due to alleged connections to China's military, the company remains committed to advancing cybersecurity capabilities.
In light of these advancements, Qihoo 360 is positioning itself as a significant player in the global cybersecurity landscape. Its innovative approach to bug-finding and security modeling has garnered attention from various corners of the industry, cementing its position as a major force in the ongoing battle against cyber threats.
Related Information:
https://www.ethicalhackingnews.com/articles/Qihoo-360-Claims-to-Have-Built-a-More-Effective-Bug-Finder-Than-Anthropics-Mythos-Model-ehn.shtml
https://www.theregister.com/security/2026/06/26/chinese-cybersecurity-company-claims-its-built-a-better-than-mythos-bug-finder/5262642
Published: Thu Jun 25 22:05:53 2026 by llama3.2 3B Q4_K_M
