AI just found another big programming language flaw

AI scanning tools have uncovered yet another critical flaw in a major piece of software, with researchers revealing a bug in Java’s memory functions which they say could allow remote code execution.

The bug, dubbed , follows the discovery of a similar flaw in the codebase of popular programming language PHP. Both were discovered by the same security research firm, Calif, which used LLMs to locate the vulnerabilities within the programming languages.

The flaw works by using malformed inputs to force a Base64 decoder within open-source Java implementation OpenJDK to overwrite neighbouring memory. This can be used to crash applications, reveal secrets within memory, or trigger remote code execution, researchers said.

Calif claimed it was able to achieve pre-authentication RCE on Atlassian Jira, Jenkins, Keycloak, and others, as well as post-auth RCE on tools such as Apache Kafka Broker and Spinnaker.

Oracle, which owns and maintains Java, has issued a critical security update for customers, but has not otherwise commented on the discovery.

The US Cybersecurity and Infrastructure Security Agency (CISA), meanwhile, also issued an advisory urging organizations to audit their IT environments for any affected Java installations and patch them as soon as possible.

For organizations that cannot patch the flaw, there are two workarounds centered on disabling the specific Base64 encoder the bug exploits. The flaw also requires an x86-64 processor with AVX-512 VBMI support, and ARM64 CPUs are not affected.

Sign up today and you will receive a free copy of our Future Focus 2026 report - the leading resource for IT decision-maker insight on priorities and investment areas in AI, security and more.

As with PHP2Shell, the discovery of this flaw is particularly significant thanks to the huge role that Java plays in global digital infrastructure.

According to TIOBE’s monthly Programming Community Index, Java is the third most popular programming language in the world and is used extensively across a range of enterprise environments.

It’s currently not known if the flaw is being exploited in the wild, although researchers from security firm Blackpoint Cyber have been able to reproduce targeted attacks which crashed a Java application, as well as extracting secrets from memory.

The incident is yet another example of AI tools uncovering a major codebase vulnerability that had gone unnoticed for more than a decade.

Calif researchers used OpenAI’s GPT-6 Cyber model, which was launched on 5 November to a select group of approved security clients in advance of wider rollout.

According to OpenAI, the model has already found 81 new vulnerabilities in Mozilla’s Firefox browser, and developed exploits for almost 950 Linux kernel flaws.

While AI companies tout the benefits of this technology for identifying new zero-day exploits, critics have warned that an explosion in the number of flaws being discovered risks overwhelming the ability of organizations to develop, deploy and apply patches for them.

Concerns have also been raised about the speed with which AI allows cyber criminals to identify and exploit security holes as part of attacks.

“It took our own Quang Luong an hour to discover the vulnerability,” Calif said as part of the announcement, “and two more hours to develop the first working exploit. The total token cost was less than $200.”

FOLLOW US ON SOCIAL MEDIA

You can also .

Original source AI just found another big programming language flaw

Back to home