Incident sheet · /security
- System
- Security & Risk
- Filed
- Read
- 3 min · 676 words
- Severity
- Routine
Databricks mitigates Java4Shell before public disclosure
Databricks disabled vulnerable code across supported runtimes before Java4Shell's disclosure. Serverless patches are automatic; classic users restart. No evidence of unauthorized access was reported.

By Daniel Okafor
What did Databricks contain before Java4Shell appeared?
By November 19 at 10 a.m. Pacific Time, Databricks had already addressed Java4Shell across every supported Databricks Runtime version, before public disclosure of the OpenJDK vulnerability.
The flaw carries identifier CVE-2026-114938. Fermín J. Serna's assessment calls it an industry-wide OpenJDK vulnerability. Databricks acted before disclosure.
Databricks stated: "We have applied the mitigation to our platform, disabling use of vulnerable code for all supported versions of Databricks Runtime."
This covers all supported versions. It names no specific runtime release number.
That supports a platform-wide mitigation claim. It does not quantify the number of clusters restarted or the operational cost.
How does mitigation differ from patching?
Databricks has organized its response into two phases. Mitigation disables vulnerable code. Patching will replace that code through patched runtime versions.
Mitigation is already active. Databricks performed it before disclosure.
To apply it as quickly as possible, the company proactively restarted affected clusters. Databricks describes that action as a rare step.
Databricks expects no meaningful performance impact from the mitigation. That is a vendor expectation, not a disclosed benchmark.
The assessment gives no latency, throughput, workload, or customer-sample data. It also gives no restart count, outage window, labor estimate, or dollar cost for the rollout.
Customers therefore have two controls: an active code disablement and a planned replacement through patched runtime versions. Databricks says the first measure blocks the attack while the second replaces vulnerable code, although its statement does not explain how the two controls differ technically.
What changes for serverless and classic clusters?
Databricks will provide patched versions of supported Databricks Runtimes. Its deployment plan differs by environment:
- Serverless: Customers receive patched versions automatically.
- Classic clusters: Customers must restart their clusters.
- Custom images: Databricks is contacting these customers directly.
The assessment gives no delivery date. It lists no affected build, fixed build, or OpenJDK version. It also omits compatibility testing, patch notes, and a rollback procedure.
Serverless users face no customer-initiated patch step in the stated plan. Classic users face a restart after patched versions become available.
Custom-image operators need to watch for direct contact. The statement does not say whether their images received the mitigation automatically. That split makes the operational burden explicit for serverless and classic environments, while leaving the treatment of custom images dependent on forthcoming customer guidance from Databricks.
Was there unauthorized access?
Databricks said: "We have no evidence to indicate any unauthorized access to Databricks products, services, or customers' data."
No compromise appears in the statement. The wording also stops short of certifying that compromise never occurred.
The assessment gives no review period, detection method, forensic coverage, log source, or customer-notification threshold. It reports no count of affected customers, exposed records, suspected attempts, or observed exploitation.
No evidence and no breach are not equivalent claims. For risk committees, the assurance remains bounded by the undisclosed review scope. For enterprise risk teams, that assurance is meaningful but incomplete, because the statement supplies neither investigative scope nor evidence capable of independent verification by customers.
What remains undisclosed?
The assessment discloses a response sequence, not a full vulnerability analysis. It omits affected OpenJDK releases, exploitation requirements, indicators of compromise, patch identifiers, and a deployment timetable.
It also omits scale metrics. No customer count, cluster count, data volume, or remediation cost appears.
The strongest conclusion available is narrow: mitigation blocked the attack before disclosure, and patching was planned. That sequence supports Databricks's containment claim. It does not establish how many systems were exposed, how long the flaw existed, or whether exploitation attempts occurred.
The company frames patching as an abundance of caution. Security teams should therefore separate two questions: Is the vulnerable code disabled, and what action will custom-image customers receive?
Serverless users should expect automatic patched versions. Classic users must plan a restart. Custom-image users need the forthcoming contact. Java4Shell's identifier and Databricks's response are clear; the assessment names neither affected versions, patch timing, remediation cost, nor the scope of its assurance review.
via databricks.com (Source)
More from Daniel Okafor
Show full bio
Staff writer covering technology and product news in enterprise information technology and the leaders who run at CIO Monitor.
205 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Java4Shell Flaw Exposes Jenkins, Jira and Keycloak to Remote Code Execution
- Java4Shell: AI found critical OpenJDK flaw in one hour, under $200
- Databricks Genie Now Runs Long Jobs — and Tells CEOs the Truth
- Microsoft ships Windows 10 KB5131096 ESU security update
- Microsoft fixes 1,235 flaws, one exploited Outlook zero-day