Java4Shell found with Codex
Partly confirmed
Importance: major (4 of 5)The takeaway
On 17 November 2026 the security startup Calif disclosed Java4Shell (CVE-2026-114938), a critical memory-corruption flaw in OpenJDK 17.0.6 and later that can allow unauthenticated remote code execution on some x86-64 servers.
Status
- Claim
Partly confirmed
- Our reporting
- Medium confidence
- Importance
- Major (4 of 5)
- Last verified
- 18 November 2026
Your AI and this story
- GPT-6.2 Astra170 days after its cutoff
- Claude Opus 5.5140 days after its cutoff
- Gemini 4 Hydrogen231 days after its cutoff
- Grok 4.7170 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 140 days before it.
Key facts
- CVE-2026-114938; CVSS 4.0 9.2 (Calif), CVSS 3.1 8.1 (Oracle/OpenJDK per SOCRadar); OpenJDK 17.0.6 and newer and builds based on it (Corretto, Temurin, Oracle Java SE); only x86-64 CPUs with AVX-512 VBMI affected, ARM64 not (Calif)
- AI role (Calif write-up by Quang Luong): about one hour with Codex to find the flaw and two more hours to a first working exploit; under $200 in tokens; useful output came after repeated prompting and filtering; ‘AI brings the horsepower. Our researchers bring the context.’
- NYT (Nov 17, via Techmeme summary; paywalled to us): Calif used GPT-6 Cyber and coordinated with Five Eyes agencies before disclosure
- Patch: OpenJDK advisory and Oracle patch released 1:00 pm PT on 17 Nov (Calif); watchTowr and Orca initially listed vendor confirmation as pending; no in-the-wild exploitation reported (ThreatDefence)
- CISA alert (17 Nov, with the Canadian Centre for Cyber Security): ‘Critical Java Vulnerability Requires Immediate Patching’: patch now, prioritise internet-facing systems, inventory bundled Java runtimes; the alert does not mention AI
- Calif also reported discovering the October PHP2Shell flaws with AI assistance (PHP2Shell: critical remote-code-execution flaws in PHP, found with GPT-5.6 Sol and Gemini 4 Argon, force emergency patching and take websites offline worldwide)
What happened
Calif named the bug after Log4Shell, the 2021 Java logging flaw, because Java runs on a very large share of enterprise servers. The disclosure was coordinated: version information came from Oracle before publication, and the advisory and patch appeared at the same time as the write-up and the NYT story. Calif’s write-up names Codex but not a model version; the NYT’s account (as summarised by Techmeme) names GPT-6 Cyber, OpenAI’s restricted cyber model.
Why it matters
It is a concrete example of the shift that national cyber agencies warned about in June: a single researcher with an AI coding agent finding a critical flaw in core infrastructure in hours. The same day, lawmakers were told that rushed patching of AI-found flaws had caused the week’s outages.
Sources
6 sources from 6 sites. Numbers match the chips in the text.
6 sources: 2 primary, 4 press
Primary
- Calif: Java4Shell: Write once, RCE everywherecalif.io, official
- CISA: Critical Java vulnerability requires immediate patchingcisa.gov, official
Press
- NYT: A.I. found a security bug in Java (Nov 17)nytimes.com, press
- Tenable: Java4Shell (CVE-2026-114938) FAQtenable.com, press
- watchTowr: Java4Shell FAQwatchtowr.com, press
- Upwind: Java4Shell: what we know nowupwind.io, press
Changes
- CISA alert
- Related PHP2Shell and APT entries
- Filed from Calif’s write-up and security-vendor FAQs; the NYT article was not readable by our fetcher; exploitation details deliberately left to the sources