Skip to content

Research · AI Engineering Study · May 28, 2026

Where the AI Pair Fails

A taxonomy of the 1,876 bugs AI-assisted engineers shipped during our 2026 benchmark. Twelve bug classes, ranked, with the code patterns each one travels on.

1,876 AI-introduced bugs · 612 engineers
Codritium Research

What we counted

A bug counts here if all four are true: (1) it shipped through the AI-assisted leg of a paired session, (2) the solo-control leg did not produce the same defect, (3) post-merge regression tests or panel review caught it within 30 days, and (4) the candidate's replay shows the AI suggestion that introduced it.

1,876

AI-introduced bugs

Across 4,200 paired sessions

44.6%

Of all defects shipped

Vs 19.1% solo baseline

12

Distinct bug classes

Mutually exclusive taxonomy

3.2

Mean lines to introduce

Surprisingly low

The taxonomy

Twelve classes cover 96% of cases. Stale-context errors — the AI proposing code against an out-of-date schema, type, or function signature — dominate by volume.

Twelve bug classes, sized by count.n=1,876 AI-introduced defects

How class shifts by language

The same AI pair fails differently by language. Stale-context errors dominate in TypeScript (fast-moving libraries, schema churn). Wrong-invariant errors spike in Go (silent error returns swallowed in suggestions). Python sees the largest share of plausible-nonsense — code that compiles and reads well but is semantically wrong.

Top-four bug classes by language.n=1,876

The patterns to watch for

Five code patterns generated 38% of all observed AI-introduced bugs. None of them are exotic — they are the patterns AI is most willing to write at a confident tone.

#PatternShareModal class
1Adapter wrapping a third-party SDK11.4%Stale context
2Generic error handler with retry8.2%Silent retry / swallow
3Permission check inside business logic6.9%Auth boundary drift
4Date / timezone normalization6.1%Wrong invariant
5Streaming response transform5.4%Race / ordering
Top-five high-yield patterns.Aggregated across all 14 codebases

What we'd change

Reviewers who learned to inspect these five patterns first cut their AI-paired regression rate by 47% in eight weeks. The class taxonomy is in the scoring rubric.