Plainly speaking, Artificial intelligence is no longer a fringe technology. It has become a core component of modern business, from customer service chatbots to complex data analysis.
We often treat the Large Language Models (LLMs) that are at the core of this technology as trusted black boxes. But like any software, they can be tampered with, manipulated, and turned against their creators. Understanding the ways an AI model can be compromised is the first step toward building a secure and resilient AI infrastructure. There are three primary ways that the authors want to cover, in which they will explain how an AI model can be compromised and made to act in unpredictable (or malicious!) ways:
Read more…
Source: Trend Micro
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- Tech support scam caused massive data breach at Australian airline Qantas
July 16, 2026
Australia’s Privacy Commissioner has revealed a tech support scam was the cause of the massive 2025 data breach at Australian airline Qantas and found the carrier didn’t breach its privacy obligations despite leaking personally identifiable information for 5.7 million customers. The Commissioner reached that conclusion, and a decision not to open a formal privacy probe, in ...
- Microsoft tops last month’s record with 622 Patch Tuesday CVEs
July 14, 2026
Remember last month when we were awed by Microsoft’s record-setting Patch Tuesday that addressed 206 CVEs? That was a quaint era compared to this month: Redmond just rolled out patches for 622 CVEs specific to its products, slightly more than tripling last month’s all-time high. Redmond’s Patch Tuesday release is once again one for the record books, with everything under ...
- Ghostcommit attack hides malicious AI instructions in images
July 13, 2026
Ghostcommit is a proof of concept that shows how AI assistants used to review software code can be tricked by hidden instructions embedded in images. The academic ASSET Research Group showed that an attacker can place instructions inside an image file, point to it in an AGENTS.md file, and get an AI coding agent to follow those instructions during a ...
- EU-UK sanctions target Russia over Europe-wide ‘vast cyber campaign’
July 13, 2026
The European Union and the UK have announced coordinated sanctions against Russia after accusing Moscow’s FSB intelligence agency of carrying out a cyber attack in December targeting Poland’s energy grid and orchestrating a wider campaign of digital sabotage across Europe. The joint measures focus on individuals and organisations linked to Russia’s security services, with the EU ...
- Trump administration subpoenas New York Times journalists over new Air Force One reporting
July 11, 2026
The Trump administration has issued subpoenas to several New York Times journalists after the newspaper reported on security concerns with the president’s new plane. The Times said its journalists were subpoenaed on Friday by the US justice department to testify before a federal grand jury in Manhattan five days later, marking the latest effort by the Trump White ...
- Ukrainians are using VPNs to cause havoc in Russia by changing fuel station statuses on maps in a bid to cause chaos and confusion
July 11, 2026
A coordinated online campaign has reportedly encouraged users to alter fuel station information on digital maps across Russia, creating confusion among drivers. The activity involves changing station statuses by marking locations with available fuel as empty or showing closed stations as operational. Supporters of the campaign claim the effort is designed to disrupt travel decisions, increase uncertainty, ...