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:
- India: Thousands of bank transfer records found spilling online after security lapse
September 26, 2025
A data spill from an unsecured cloud server has exposed hundreds of thousands of sensitive bank transfer documents in India, revealing account numbers, transaction figures, and individuals’ contact details. Researchers at cybersecurity firm UpGuard discovered in late August a publicly accessible Amazon-hosted storage server containing 273,000 PDF documents relating to bank transfers of Indian customers. Read more… Source: ...
- XCSSET evolves again: Analyzing the latest updates to XCSSET’s inventory
September 25, 2025
Microsoft Threat Intelligence has identified yet another XCSSET variant in the wild that introduces further updates and new modules beyond those detailed in our March 2025 blog post. The XCSSET malware is designed to infect Xcode projects, typically used by software developers, and run while an Xcode project is being built. We assess that this mode ...
- CISA Directs Federal Agencies to Identify and Mitigate Potential Compromise of Cisco Devices
September 25, 2025
Today, CISA issued Emergency Directive ED 25-03: Identify and Mitigate Potential Compromise of Cisco Devices to address vulnerabilities in Cisco Adaptive Security Appliances (ASA) and Cisco Firepower devices. CISA has added vulnerabilities CVE-2025-20333 and CVE-2025-20362 to the Known Exploited Vulnerabilities Catalog. The Emergency Directive requires federal agencies to identify, analyze, and mitigate potential compromises immediately. Agencies ...
- New LockBit 5.0 Targets Windows, Linux, ESXi
September 25, 2025
Trend™ Research has identified and analyzed the source binaries of a new LockBit version in the wild, which is the latest from the group’s activities following the February 2024 law enforcement operation (Operation Cronos) that disrupted their infrastructure. In early September, the LockBit ransomware group reportedly resurfaced for their sixth anniversary, announcing the release of “LockBit ...
- Co-op digests £80m profit hit from cyber attack
September 25, 2025
he Co-operative Group has revealed an £80m hit to half-year profit after a damaging cyber attack earlier this year. The retail and funerals specialist said that it had slumped to a loss in the first half of 2025 after being targeted by a “malicious” cyber attack. Shoppers were faced with empty shelves and issues with payments ...
- Another BRICKSTORM: Stealthy Backdoor Enabling Espionage into Tech and Legal Sectors
September 24, 2025
Google Threat Intelligence Group (GTIG) is tracking BRICKSTORM malware activity, which is being used to maintain persistent access to victim organizations in the United States. Since March 2025, Mandiant Consulting has responded to intrusions across a range of industry verticals, most notably legal services, Software as a Service (SaaS) providers, Business Process Outsourcers (BPOs), and Technology. ...