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

What happened
Researchers tested seven AI models launched since February 2026 to see if they could find a newly reported security flaw in Google's V8 engine. In September, the GPT 6 Sol model identified the vulnerability in 13 minutes for a cost of $4.24. Earlier in the year, the same task took 25 hours and $350 with less advanced models, and $22 in 44 minutes with GPT 5.6 Sol in July. Only three of the seven models were able to detect the flaw, while the others failed.
Key details
- The vulnerability targeted was CVE-2026-102321, a type confusion bug in Google V8.
- GPT 6 Sol found the bug in 13 minutes for $4.24.
- GPT 5.6 Sol and Claude Opus 5 also found the bug, with Claude Opus 5 costing $43.40.
- DeepSeek and Z.ai models did not find the vulnerability.
- Token prices for AI models have dropped by about 47% per quarter since 2023.
Why it matters
Cybersecurity teams, companies, and attackers are affected by the sharp reduction in cost and time needed to find zero-day vulnerabilities using AI. This means that individuals with less expertise and fewer resources can now use advanced AI models to identify security flaws, increasing the risk of software breaches. Organizations now face the need to update their security practices and may need to use similar AI tools themselves to detect vulnerabilities before they are exploited.
