Tag: LLM
-
Quarkslab Says Anti-Reversing Software Should Return Plausible Wrong Answers Instead of Crashing
Quarkslab has argued that LLM-assisted reverse engineering does not make obfuscation obsolete, but it changes the defender’s threat model. Its latest experiment found that autonomous coding agents routinely avoid difficult deobfuscation, pivot to dynamic analysis, and often stop once they obtain an answer that appears credible even if it is wrong. The research firm tested…
-
Black Hat 2026: What should you be allowed to talk to AI about? FireTail Blog
Tags: ai, attack, business, conference, control, cybersecurity, data, detection, edr, framework, LLM, strategy, technology, tool, vulnerability, vulnerability-managementAug 21, 2026 – Jeremy Snyder – If you were at RSA Conference last year, you probably remember the goats. Or the puppies. Or the miniature petting zoos. It was a year of “over-the-top” spectacle. A bit of a circus, if I’m being honest.Coming into RSAC 2026, the vibe shifted. The show floor was noticeably…
-
Bad Grammar is Dead. Tone Matching is the New Payload
Generative AI has eliminated typos in phishing. Discover how attackers weaponize LLMs for perfect tone matching, and how Doppel stops them at machine speed. First seen on securityboulevard.com Jump to article: securityboulevard.com/2026/08/bad-grammar-is-dead-tone-matching-is-the-new-payload/
-
Why MSPs Need Visibility Across ‘Every AI Surface’: KIPIO CEO
Amid the AI adoption rush, the usage of LLM-powered tools in unsanctioned or insecure ways poses a massive risk for MSPs and their clients, according to KIPIO co-founder and CEO Emily Hock. First seen on crn.com Jump to article: www.crn.com/news/security/2026/why-msps-need-visibility-across-every-ai-surface-kipio-ceo
-
Grok exfiltrates user data when malicious instructions are encrypted
Cryptographic Context Injection is only the latest way to break an LLM safety guardrail. First seen on arstechnica.com Jump to article: arstechnica.com/security/2026/08/grok-exfiltrates-user-data-when-malicious-instructions-are-encrypted/
-
AI Security Incident Case: Encrypted Reasoning Blocks of Proprietary LLMs Can Be Stolen via Cross-Model Replay
Overview On August 10, 2026, MATS Research, the ELLIS Institute Tübingen, the Max Planck Institute for Intelligent Systems, and other institutions jointly published the paper Stealing Reasoning Traces from Proprietary LLM APIs. The research reveals a common architectural flaw in the reasoning model APIs of three major AI providers, Anthropic, OpenAI, and Google, which allows……
-
LLMs and Contextual Integrity
I have been thinking a lot about AI and integrity. Part of that is contextual integrity. I recently found two papers on the topic. “CIMemories: A Compositional Benchmark for Contextual Integrity of Persistent Memory in LLMs”: Abstract: Large Language Models (LLMs) increasingly use persistent memory from past interactions to enhance personalization and task performance. However,…
-
Adam Shostack Talks Hugging Face & PHANTOM-B
World-class threat modeler Adam Shostack shared he was blown away by OpenAI’s revelations about the Hugging Face attack, and explains why his new threat model for LLMs is both lightweight yet still usable. First seen on darkreading.com Jump to article: www.darkreading.com/vulnerabilities-threats/adam-shostack-talks-hugging-face-phantom-b
-
Prompt injection remains top LLM threat, OWASP report finds
First seen on scworld.com Jump to article: www.scworld.com/brief/prompt-injection-remains-top-llm-threat-owasp-report-finds
-
AI Recommendation Poisoning: How “Ask AI” Buttons Silently Alter LLM Memory
A new class of prompt injection is spreading across commercial websites. It requires no malware, no stolen credentials, and no zero-day exploit. It abuses a standard feature built into almost every major AI assistant: pre-filled deep links.We observed production websites embedding hidden prompt injection payloads inside “Ask AI” buttons on marketing and competitor comparison pages.…
-
South Korea’s government overtakes telcos as top cyber attack target
Kaspersky researcher Sojun Ryu says ransomware crews have joined nation-state groups in going after South Korean organisations, as traces of LLM output start turning up inside malware First seen on computerweekly.com Jump to article: www.computerweekly.com/news/366647735/South-Koreas-government-overtakes-telcos-as-top-cyber-attack-target
-
OWASP 2026 LLM Top 10: >>The model will be fooled<<
The OWASP GenAI Security Project has released the 2026 edition of its Top 10 for LLM Applications and, for the first time, the list was influenced by real-world incidents. The … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/06/owasp-2026-llm-top-10-released/
-
AI Sends Global Crime Syndicates Into Fraud Nirvana
Organized crime is convincingly scamming at scale, making billions thanks to AI-enabled voice cloning, deepfake real-time video overlays, LLM-driven persona management, and automated translation. First seen on darkreading.com Jump to article: www.darkreading.com/threat-intelligence/ai-global-crime-syndicates-fraud-nirvana
-
Metasploit Opens Its Exploit Engine to LLMs via New MCP Integration
If there was any reason to patch your systems, this would be it: The release of Metasploit 6.5, which brings the penetration testing framework into the AI age. Now, script kiddies and sophisticated foreign operatives don’t even have to cut and paste their code to break into your systems. They can just ask Metasploit to..…
-
Prompt Injection Remains Biggest LLM Risk, Despite Limited Incidents
Prompt injection remains the most dangerous security threat to LLMs, according to OWASP’s latest Top 10 LLM Applications list First seen on infosecurity-magazine.com Jump to article: www.infosecurity-magazine.com/news/prompt-injection-llm-risk/
-
Six Flowise Vulnerabilities Enable Remote Code Execution on AI Workflow Servers
Six newly disclosed vulnerabilities in Flowise, a popular open”‘source platform for building AI agents and LLM workflows, allow unauthenticated and low”‘privileged attackers to achieve remote code execution (RCE) on self”‘hosted and cloud AI workflow servers running vulnerable versions. These flaws collectively expose organizations to full server compromise, data exfiltration, and AI pipeline manipulation if instances…
-
AI Agent Guardrails: How to Set Boundaries Before You Give an LLM Access to Your Systems
AI agents are crossing a line that traditional chatbots never crossed. A chatbot produces text. An AI agent can retrieve customer records, query financial data,…Read More First seen on securityboulevard.com Jump to article: https://securityboulevard.com/2026/08/ai-agent-guardrails-how-to-set-boundaries-before-you-give-an-llm-access-to-your-systems/
-
Chinese hacker used DeepSeek to launch autonomous cyberattacks on vulnerable servers
A Chinese threat actor operating under the aliases >>knaithe<>KnYuan<< used multiple LLMs to automate cyberattacks against internet-facing systems … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/03/deepseek-ai-autonomous-cyberattacks-hermes-agent/
-
30 days with Claude Mythos Preview: How Tenable adapted our security program, and why yours is next
Tags: ai, api, attack, business, control, cybersecurity, data, data-breach, endpoint, exploit, flaw, injection, LLM, remote-code-execution, risk, service, threat, tool, update, vulnerabilityTenable spent 30 days running frontier AI models against our own code. It didn’t just find bugs, it proved they’re real, with reproducible exploits. That fundamentally changes code security from ranking potential code defects to a much higher signal focused on the findings that matter. Read on to learn how it reshaped our security team’s…
-
Klassische Firewalls sind blind für Angriffe in natürlicher Sprache – Warum LLM-Guards allein die KI-Sicherheit nicht retten
First seen on security-insider.de Jump to article: www.security-insider.de/llm-guards-ki-sicherheit-a-2fbccecb3c2d06bb933157bbaadd9970/
-
What an LLM Can Find: A Practical, Cheap Path to Code-level Threat Discovery
An AI-assisted audit found 29 flaws in GlobaLeaks, showing LLMs make large-scale code reviews faster, cheaper, and accessible. GlobaLeaks, a mature whistleblowing platform that had already undergone six independent professional audits over the past thirteen years, was subjected to an LLM-assisted security review that cost roughly USD 3,140 in API calls. The review identified 29…
-
(g+) Agentgateway: Ein Kontrollpunkt für LLMs, Tools und KI-Agenten
Je mehr Werkzeuge KI-Agenten nutzen, desto schwieriger wird es nachzuvollziehen, wer worauf zugreift. Agentgateway soll genau dieses Problem lösen. First seen on golem.de Jump to article: www.golem.de/news/agentgateway-ein-kontrollpunkt-fuer-llms-tools-und-ki-agenten-2607-211089.html
-
Google says it fixed more Chrome bugs in June than over the past two years, thanks to AI
As experts have warned for the last two years, some companies, like Microsoft and now Google, are finding and patching an exponential number of bugs in their products, thanks to the use of LLMs and AI tools. First seen on techcrunch.com Jump to article: techcrunch.com/2026/07/30/google-says-it-fixed-more-chrome-bugs-in-june-than-over-the-past-two-years-thanks-to-ai/
-
Stronger AI Safety Requires Peeking Inside the ‘Black Box’
Researchers propose focusing on identification of certain cognitive elements in LLMs that indicate when AI systems may take an unwanted action. First seen on darkreading.com Jump to article: www.darkreading.com/cybersecurity-analytics/stronger-ai-safety-requires-peeking-inside-black-box
-
Using LLMs to Find & Prioritize Vulnerabilities Is No Easy Task
The latest large language models have high false-positive rates and fail to take into account the context of scans, leading to more work for AppSec professionals. First seen on darkreading.com Jump to article: www.darkreading.com/application-security/finding-and-prioritizing-vulnerabilities-no-easy-task
-
Using LLMs to Find & Prioritize Vulnerabilities Is No Easy Task
The latest large language models have high false-positive rates and fail to take into account the context of scans, leading to more work for AppSec professionals. First seen on darkreading.com Jump to article: www.darkreading.com/application-security/finding-and-prioritizing-vulnerabilities-no-easy-task
-
Die neue Risiken durch agentische KI im Fokus – Warum LLM-Observability zum Sicherheitsfaktor für KI-Systeme wird
First seen on security-insider.de Jump to article: www.security-insider.de/warum-llm-observability-zum-sicherheitsfaktor-fuer-ki-systeme-wird-a-bd9def5d1852e535db0558409c91e25f/
-
When AI Attacks: OpenAI Models Autonomously Hack Hugging Face
Advanced LLMs escaped their sandboxes while attempting to achieve a non-malicious benchmark test objective. First seen on darkreading.com Jump to article: www.darkreading.com/cyber-risk/openai-models-autonomously-hack-hugging-face