AI is now finding bugs in the code that runs banks, airlines, and hospitals

Skye Jacobs

Posts: 2,307   +64
Staff
The takeaway: An AI model built for security research has uncovered a series of vulnerabilities in Java, the Oracle programming language used in the systems of airlines, banks, hospitals, and government agencies. The flaws were discovered by cybersecurity startup Calif using OpenAI's GPT-6 Cyber model. Now, the Cybersecurity and Infrastructure Security Agency urged companies to update every system running Java.

Oracle released what it called a "critical security patch update" the same day and urged customers to install it. The company declined to comment further. Applying the fix could be a major undertaking: 90% of Fortune 500 companies use Java, which is deeply embedded in the software that keeps large organizations running.

Bruce Dang, a senior researcher at Calif, said the language's widespread use is what makes the discovery so serious. "Java is basically critical infrastructure," he told The Financial Times. "It's so central to the technology that we depend on . . . [and] a lot of big companies are going to have their work cut out to get it fixed."

Calif said it notified Oracle about the vulnerability last month. It then worked with government agencies through the Five Eyes intelligence-sharing alliance, which includes Australia, Canada, New Zealand, the UK and the US. Amazon, Apple, Google, and Microsoft also helped coordinate fixes. That kind of cooperation is intended to ensure patches are ready before details of a vulnerability spread widely enough for attackers to exploit it.

The Java findings are the latest in a growing list of software vulnerabilities discovered with the help of AI models in recent months. In October, researchers used models from OpenAI and Google to identify two security flaws in PHP, a programming language used by 70% of websites. The bugs had gone undetected for more than a decade. After they were disclosed, companies and government agencies took websites offline as a precaution.

These cases illustrate a shift in how vulnerability research is conducted. AI models can now analyze large, established codebases and uncover flaws that years of human review have missed. That gives defenders an advantage, but it also puts them in a race against potential attackers.

Security researchers are trying to identify and patch weaknesses before malicious actors can exploit them. Advances in AI make that race more challenging because the same capabilities could help attackers discover the same vulnerabilities.

The CISA advisory comes during a period of unusual instability affecting digital services in the US, adding to concerns about further disruptions. Microsoft experienced a widespread outage last week that disrupted its services. The company said a system malfunctioned during a routine effort to reroute user traffic and ruled out a cyberattack. JPMorgan and United Airlines are among the other companies that have experienced outages in recent weeks.

Air travel has also been affected. A five-hour disruption to the country's aging air traffic control system caused a weekend of travel chaos for American passengers. Separately, Norwegian oil and gas group Equinor said it had been targeted in a cyberattack.

US Secretary of State Marco Rubio sought to ease concerns that the outages were connected. Speaking to reporters in Manila during President Donald Trump's three-day trip to Asia, he rejected the idea of a coordinated campaign. "We are not under systemic attack. These events, one has nothing to do with another, they are not interrelated, they are all separate and distinct," he said.

"Every indicator we have is that every single one of these events has an explanation or a cause that has nothing to do with a cyber criminal, has nothing to do with a hostile actor or another country," Rubio added.

The outages have not been linked to the Java vulnerabilities. Still, the timing puts additional pressure on organizations that rely heavily on Java. With Oracle's patch released and CISA's warning issued, companies using Java must now identify every place the language is embedded in their systems and apply the necessary updates.

Permalink to story:

 
Ai is "fixing" bugs...but is A.I leaving any sort of footprint that another A.I.model could exploit?

Or when companies use A.I to write code, how do we know they are not leaving behind some type of backdoor for their own self preservation.

I wouldn't trust A.I writing the country's security and defense algorithms not truly knowing what they are leaving behind harder for humans to identify.

That's when "fixing" the bugs becomes the Trojan horse.
 
Ai is "fixing" bugs...but is A.I leaving any sort of footprint that another A.I.model could exploit?

Or when companies use A.I to write code, how do we know they are not leaving behind some type of backdoor for their own self preservation.

I wouldn't trust A.I writing the country's security and defense algorithms not truly knowing what they are leaving behind harder for humans to identify.

That's when "fixing" the bugs becomes the Trojan horse.
How do you know the humans coders were not doing that?

If you don't trust AI code you shouldnt be trusting human code either.
 
Oh and now that the word is out, hackers throughout the world are going to be trying to find a way IN before they fix it.
 
How do you know the humans coders were not doing that?

If you don't trust AI code you shouldnt be trusting human code either.

Equation group intercepted hardware mid shipment, planted tiny remot backdoors and keyloggers in all kinds of crap, then resumed the shipment to the buyers in all sectors. then used those backdoors to supply chain exploit code repositories of target companies to plant backdoors..

so yes, humans are bad, AI is bad.. it is all bad.
 
Not sure why Java would be difficult to fix? It wasn't a language I used a great deal but it was no harder to code or fix bugs in than any other?
 
Back