AI tools accelerate software vulnerability exploitation

Artificial intelligence is significantly compressing the window for exploiting software vulnerabilities. According to the 2026 Global Threat Report by FortiGuard Labs, automated offensive tools have reduced the time required to validate and exploit flaws to between 24 and 48 hours, a sharp decline from the historical average of approximately one week.

In Latin America, the region saw nearly 15,000 million exploitation attempts in 2025, representing a nearly 40 percent increase. This surge is attributed to automated routines that can chain together multiple low- or medium-severity flaws to create high-severity breaches.

Defensive efforts face a significant asymmetry. While AI-driven detection can achieve high precision in discovering complex bugs, remediation remains a challenge. Current frontier models often refuse to execute remediation prompts due to built-in guardrails, and existing automated patching attempts frequently miss secondary issues or break production code. The persistence of network intrusions is largely linked to delayed patch deployment and unadjusted system configurations rather than the novelty of the flaws themselves.