AI hit 65% of committed code a year early. Do we know what's in the box?
SonarSource’s survey of 1,026 developers found AI-assisted code generation accelerated faster than expected, reaching 65% of committed code by 2026—a year ahead of 2025 predictions—while verification practices lag dangerously: 76% report generating AI code faster than they can verify it, and the share who consistently check before committing dropped from 48% to 39% despite 94% having incomplete trust in AI code quality. Security concerns top developer worries, with 61% extremely concerned about AI-powered attacks finding and exploiting vulnerabilities at scale. Teams employing inline verification during code generation (rather than at PR review) and well-defined governance policies report 31–32% better outcomes on defect and patch costs, suggesting that late-stage review is insufficient for the autonomy developers are deploying into business-critical systems.