Cybersecurity startup Depthfirst claimed its AI model discovered critical internet bugs missed by Anthropic’s Mythos system at significantly lower cost.
SAN FRANCISCO: Cybersecurity startup Depthfirst has claimed its artificial intelligence system identified major internet security flaws missed by Anthropic’s recently launched AI model Mythos, intensifying competition in the growing AI-driven cybersecurity sector. The company said its technology was able to detect additional critical bugs affecting widely used internet infrastructure while operating at a fraction of the reported cost.
Anthropic introduced Mythos last month, describing the platform as an advanced cybersecurity AI capable of uncovering severe vulnerabilities in critical internet code. The announcement drew widespread attention across the technology and cybersecurity industries because of the model’s ability to identify previously unknown software flaws that could potentially impact millions of online users.
However, Depthfirst chief executive Qasim Mithani now claims his company’s specialised AI system discovered even more vulnerabilities than Mythos while using significantly fewer computing resources. According to Mithani, the startup’s approach focuses on optimising AI models for highly specific cybersecurity tasks instead of broader general-purpose functions.
The company stated its system can reportedly achieve results costing around US$1,000 compared with an estimated US$10,000 required for similar analysis using Anthropic’s technology. Industry analysts say the development highlights growing competition among cybersecurity companies seeking to build more efficient AI-powered threat detection systems.
Experts believe AI-driven vulnerability discovery could transform global cybersecurity by identifying hidden weaknesses faster than traditional manual testing methods. At the same time, specialists warn that increasingly powerful AI systems may also create new risks if advanced vulnerability detection tools fall into malicious hands.
The startup did not publicly disclose full technical details about the newly discovered flaws but claimed some vulnerabilities could affect a large portion of internet users through widely used software and infrastructure systems.
The latest AI cybersecurity competition also reflects broader industry efforts to commercialise artificial intelligence tools capable of automating complex digital defence operations. Analysts expect investment in AI-powered cyber protection systems to accelerate further as companies and governments respond to rising online security threats worldwide.
