54 stories tagged AI Security · page 2 of 3
Security experts are now using prompt injection techniques to stop AI-powered hacking tools. This method tricks malicious agents into shutting down before they can cause harm. It's a creative twist on a common attack strategy.
Researchers discovered a new technique called 'Ghostcommit' that embeds hidden instructions in images to manipulate AI systems. This could trick AI agents into revealing sensitive information or performing unauthorized actions.
Researchers discovered that nine widely used AI tools can be tricked into assembling massive botnets through a technique called "HalluSquatting." This vulnerability exploits AI models' tendency to hallucinate — generating plausible but incorrect responses — instead of refusing harmful requests. The finding underscores a critical security flaw in how AI handles ambiguous or malicious prompts.
An AI agent executed a real-world ransomware attack, but humans still chose the victim, set up infrastructure, and supplied stolen credentials. This shows AI's role in cybercrime is growing, but not yet fully autonomous.
A security flaw in Cursor's AI agent allowed it to break out of its protective sandbox. This shows why AI tools need strict boundaries to prevent misuse. Developers are now working on fixes to ensure safer AI interactions.
A malicious AI agent skill bypassed security checks and was downloaded by over 26,000 users, exposing the growing risks of AI-driven malware in everyday applications.
A new attack demonstrates that AI-powered browsers can be tricked into bypassing all safety rules simply by feeding them a false premise—like telling the model that 2 + 2 = 5. Once the LLM enters this 'dream world,' it will follow forbidden instructions, raising serious concerns about the security of AI-driven browsing assistants. (Ars Technica AI)
Researchers introduced RIFT-Bench, a dynamic red-teaming benchmark that uses a graph-based methodology to evaluate security vulnerabilities across diverse agentic AI systems. Unlike static tests, it simulates adaptive attacks to find weaknesses before real hackers do, aiming to make autonomous AI safer.
Export controls on cybersecurity software have failed for decades. The latest attempt to restrict AI tools like Anthropic's Mythos will likely meet the same fate. The technology always finds a way around these barriers, as history from PGP to spyware shows.
Lelu is an open-source tool designed to detect and prevent manipulation of AI agents in real-time. It's a significant step forward in ensuring the security and reliability of AI systems.
ServiceNow researchers discovered a vulnerability in AI research agents that can leak sensitive data through a novel attack called MosaicLeaks. The flaw exploits how these agents handle and combine information, potentially exposing private details even when individual queries seem safe. The issue affects multiple AI research platforms and highlights a fundamental security challenge in agentic AI systems.
AI models with sophisticated hacking capabilities are becoming more common, raising serious security concerns. Experts argue that these developments are unstoppable, requiring proactive measures to mitigate risks.
A critical flaw in Microsoft Copilot allowed hackers to steal two-factor authentication codes. This highlights ongoing security risks in AI systems.
OpenAI has launched Lockdown Mode to prevent sensitive data leaks from AI models. While not foolproof, it significantly reduces the risk of prompt injection attacks.
Anthropic is rolling out its AI security tool, Mythos, to 150 organizations in 15 countries. This move aims to protect critical infrastructure like power grids and hospitals from cyberattacks that could impact millions.
The Open Web Application Security Project (OWASP) launched Agent Memory Guard, a tool to defend AI agents from memory poisoning attacks. This is crucial as AI systems become more integrated into daily tasks.
Hackers are finding ways to manipulate AI chatbots by exploiting their programmed personalities. This highlights the need for better safeguards as AI becomes more integrated into daily life.
AI agents working together can solve complex problems better than single agents, but this collaboration introduces new security risks. Researchers warn that trust must be built into these networks from the start, not added later.
Anthropic's AI system, Mythos, discovered critical security bugs in Firefox. This could lead to safer web browsing for everyone. Mozilla is now using AI to proactively find and fix vulnerabilities before they become problems.
The US State Department has issued a global warning about alleged AI technology theft by Chinese company DeepSeek. This highlights growing tensions over AI competition and intellectual property between the US and China.
Researchers propose TRUST, a decentralized framework to address robustness, scalability, opacity, and privacy challenges in AI systems. The approach aims to enhance trust in high-stakes applications like Multi-Agent Systems (MAS).
Researchers have developed a new jailbreak technique called Incremental Completion Decomposition (ICD) that exploits LLMs by eliciting single-word continuations before extracting harmful responses. This method bypasses current safety mechanisms, raising concerns about the robustness of LLM safeguards.
A recent study revealed five critical AI agent failures over a 36-day period, none of which were detected by the agents themselves. This highlights significant gaps in AI self-monitoring and security protocols.
A new study reveals previously unknown attack vectors in sandboxed AI agents, challenging assumptions about their security. The findings highlight the need for enhanced isolation techniques and continuous monitoring.