What happened

On Wed, 05 Aug 2026 02:10:00 UTC, the latest version of Neo4j core product was rolled out to the Aura Environment. This release contained a bug in the Cypher trim() family of functions (trim, ltrim, rtrim, btrim) that incorrectly handled string length encoding during write operations.

The defect could produce two outcomes when a query using these functions modified stored string data:

• Query failure — the operation would error and no data was written.
• Silent data issue — the operation would succeed, but the stored string's length metadata could be recorded incorrectly, potentially resulting in trailing whitespace padding.

The issue was identified on Wed, 05 Aug 2026 17:51:51 UTC after alerting surfaced read-back failures on a single free-tier database. Our engineering team immediately reverted the introducing code change and began preparing a hot-fixed release. The patched version was rolled out across the affected tiers by Fri, 07 Aug 2026 16:02:49 UTC, and an on-premises patch was also published. A customer-facing status page was maintained throughout at https://status.neo4j.io/incidents/kjx6l7442j2v.

How the service was affected

The impact was limited to Aura databases running Neo4j version 2026.07.0-8 that executed write queries using trim()-family Cypher functions during the exposure window.

• Query failures: A small but measurable increase in query errors was observed for transactions involving trim(). These failed queries did not persist any data changes.
• Trailing whitespace: In some cases, successful writes stored string values with incorrect trailing whitespace. Through systematic scanning, our team identified a number database instances that showed evidence of this issue.
• Unreadable properties: In one case, a property on a free-tier database instance became unreadable. The systematic scan established that this only occurred once across the whole platform.

Log analysis determined that a small number of database instances across several organizations may have executed potentially affected write queries.

Our engineering team proactively contacted administrators and owners of affected organizations, provided per-database instance diagnostic reports offering guidance on identifying and remediating impacted data. No widespread unreadable-data corruption was found during the fleet-wide consistency and whitespace scans conducted after the fix was deployed.

What we are doing now

We take data integrity seriously and have undertaken a thorough response to ensure affected customers are supported and to reduce the likelihood of similar issues in the future. Our efforts fall into the following areas:

• Immediate fix — The code change responsible for the defect was reverted and a patched release was deployed across all affected Aura tiers. An on-premises patch was also published for self-managed users.
• Proactive customer outreach — We identified potentially affected databases through log analysis and ran two rounds of offline scans: consistency checks for unreadable-data corruption, and a purpose-built scanner for trailing-whitespace issues. Per-database diagnostic reports were generated and made available, and affected customers were directly notified with guidance.
• Detection improvements — We are evaluating enhanced monitoring for per-function error rates in query processing, which could help surface similar issues earlier during staged rollouts.
• Rollout process review — We are examining our rollout procedures, including whether to introduce additional triage windows between rollout phases and how to prevent databases from inadvertently advancing to newer versions during operations like pause/resume.
• Ongoing follow-ups — Unrelated storage inconsistencies discovered during scans have been tracked and are being addressed by our engineering team through separate work streams.