AI-Assisted Development Outpaces Security Review and Release Governance
Emerging Technology Risk · 2 sources · Updated 4d ago · First seen Nov 18, 2026

AI-assisted development is increasing software delivery volume faster than security teams can assess changes and document release decisions, according to Chris Hughes’s Resilient Cyber article citing Cytix-commissioned research. Among 250 UK security leaders surveyed, 78% reported mandated AI-assisted development, but only 38% strongly agreed their organizations were prepared for the resulting code volume. Nearly half had approved releases without full confidence in their security, while 53% said findings were frequently outdated before reaching those responsible for acting on them. GitHub’s Octoverse report illustrates the scale of development activity: developers pushed code 17.9 billion times over the preceding 12 months, and agents initiated more than 40% of pull requests by September, up from 17% in March.
The findings point to a broader software-governance gap, not solely vulnerabilities in AI-generated code. Reported incident causes included dependencies, out-of-process changes, misconfigurations, longstanding bugs, AI-generated weaknesses, and unreviewed changes. GitHub also found that new repositories containing agent configuration files averaged eight times as many commits as those without them, an association that underscores review pressure without establishing causation. Hughes advocates timely, change-specific security assessments, scrutiny proportionate to risk, explicit risk acceptance, and continuously collected evidence connecting testing and approval decisions to individual changes.


