AI finds zero-day flaws for just $4 now – even mediocre hackers can cash in
Original reporting by ZDNet

The ability of artificial intelligence models to discover zero-day vulnerabilities—previously unknown software flaws—is rapidly advancing, becoming both more effective and dramatically cheaper. This progress, fueled by intense competition and falling token prices (down 47% quarterly since 2023), presents a profound double-edged sword for cybersecurity. While AI offers unprecedented tools for business innovation, it simultaneously lowers the barrier for attackers, making sophisticated AI-assisted breaches more accessible and affordable. This escalating capability transforms the landscape of digital defense, demanding an urgent reevaluation of organizational security postures.
The $4 Breakthrough
Recent research by Xint.io starkly illustrates this shift, chronicling the evolution of AI's hacking prowess in 2026. The study tasked seven leading AI models with finding a critical type confusion bug (CVE-2026-102321) in Google’s V8 engine—a vulnerability unknown to their training data. In February, even the best models failed after 25 hours and $350. Yet, by July, OpenAI's GPT 5.6 Sol successfully located the flaw in just 44 minutes for $22. The most striking development came in September, when GPT 6 Sol identified the zero-day with its first guided prompt in a mere 13 minutes, costing only $4.24. Only three of the seven tested models ultimately succeeded, highlighting varying capabilities. This dramatic acceleration and cost reduction underscore a new reality: what was once too complex and expensive for many attackers is now within reach, pushing defenders to innovate at an equally rapid pace by leveraging these same AI tools.
The convergence of rapidly advancing AI models and plummeting computational costs is fundamentally reshaping the cybersecurity landscape. The research unequivocally demonstrates AI's accelerating capability to discover zero-day vulnerabilities in minutes and for mere dollars, a feat once reserved for highly specialized human experts. This pivotal shift democratizes sophisticated cyber exploitation, enabling a broader range of threat actors to leverage cutting-edge tools, thus dramatically escalating the overall risk profile for organizations globally. The era where prohibitive costs and elite expertise served as primary barriers to entry for advanced hacking is rapidly receding, demanding immediate strategic adaptation.
The Evolving Security Imperative
This new reality, while empowering attackers, simultaneously presents an unprecedented opportunity for defenders. Organizations are now compelled to integrate AI-driven tools into their security operations, proactively scanning for vulnerabilities and hardening their digital perimeters with unparalleled speed and efficiency. This necessitates a profound recalibration, elevating AI-assisted security from a future consideration to an immediate, non-negotiable imperative. However, human expertise remains indispensable; not in the brute-force task of scanning, but in intelligently guiding AI, interpreting its nuanced findings, and designing robust, resilient security architectures rooted in foundational principles like zero-trust. The future of cybersecurity will be characterized by an accelerating, AI-powered arms race, where sustained success will hinge on an organization's agility to harness AI's dual capacity – both to anticipate and neutralize emergent threats, and to continuously fortify defenses against an increasingly capable adversary.
Frequently asked questions
- How is AI technology changing the landscape of discovering software vulnerabilities and cyberattacks?
- AI is rapidly advancing in its ability to identify complex software vulnerabilities, including previously unknown "zero-day" flaws. Concurrently, the cost of utilizing AI models for such tasks is dramatically decreasing. This dual development makes sophisticated vulnerability discovery more accessible, presenting both opportunities for enhanced defensive security measures and new challenges from potential adversaries leveraging AI for cheaper, more effective cyberattacks.
- Which specific AI models are most effective at identifying zero-day software vulnerabilities currently?
- Recent research indicates that advanced large language models are proving effective at identifying zero-day vulnerabilities. Specifically, models such as GPT 5.6 Sol, GPT 6 Sol, and Claude Opus 5 have demonstrated the capability to discover critical software flaws that were previously unknown. These models achieve this with varying efficiency and cost, showcasing a significant leap in AI's ability to perform complex security analysis tasks.
- How does the falling cost of AI tools affect cybersecurity defense strategies?
- The decreasing cost of AI tools presents a dual challenge and opportunity for cybersecurity defense. While it lowers the barrier for adversaries to launch sophisticated AI-assisted attacks, organizations can also leverage these cheaper tools to proactively identify vulnerabilities in their own systems. However, human expertise remains crucial for effectively guiding AI models and implementing comprehensive security, as AI augments rather than replaces traditional security practices like zero-trust and network segmentation.