EchoGram Flaw Bypasses Guardrails in Major LLMs
HiddenLayer reveals the EchoGram vulnerability, which bypasses safety guardrails on GPT-5.1 and other major LLMs, giving security teams just a 3-month head start.
ai
More Prompt||GTFO
The next three in this series on online events highlighting interesting uses of AI in cybersecurity are online: #4, #5, and #6. Well worth watching.
Critical RCE Flaws in AI Inference Engines Expose Meta, Nvidia, and Microsoft Frameworks
Security researchers at Oligo Security have uncovered a series of critical Remote Code Execution vulnerabilities affecting widely deployed AI inference servers from major technology companies. The flaws affect frameworks developed by Meta, NVIDIA, Micr…
Anthropic: China-backed hackers launch first large-scale autonomous AI cyberattack
China-linked actors used Anthropic’s AI to automate and run cyberattacks in a sophisticated 2025 espionage campaign using advanced agentic tools. China-linked threat actors used Anthropic’s AI to automate and execute cyberattacks in a highly sophisticated espionage campaign in September 2025. The cyber spies leveraged advanced “agentic” capabilities rather than using AI only for guidance. Attackers […]
Chinese State Hackers Jailbroke Claude AI Code for Automated Breaches
Anthropic, the developer behind Claude AI, says a Chinese state sponsored group used its model to automate most of a cyber espionage operation against about 30 companies with Claude handling up to 90% of the technical work.
The Role of Humans in an AI-Powered World
As AI capabilities grow, we must delineate the roles that should remain exclusively human. The line seems to be between fact-based decisions and judgment-based decisions.
For example, in a medical context, if an AI was demonstrably better at reading a test result and diagnosing cancer than a human, you would take the AI in a second. You want the more accurate tool. But justice is harder because justice is inherently a human quality in a way that “Is this tumor cancerous?” is not. That’s a fact-based question. “What’s the right thing to do here?” is a human-based question…
Germany’s BSI issues guidelines to counter evasion attacks targeting LLMs
Germany’s BSI warns of rising evasion attacks on LLMs, issuing guidance to help developers and IT managers secure AI systems. Germany’s BSI warns of rising evasion attacks on LLMs, issuing guidance to help developers and IT managers secure AI systems and mitigate related risks. A significant and evolving threat to AI systems based on large […]
Anthropic Says Claude AI Powered 90% of Chinese Espionage Campaign
A state-sponsored threat actor manipulated Claude Code to execute cyberattacks on roughly 30 organizations worldwide.
The post Anthropic Says Claude AI Powered 90% of Chinese Espionage Campaign appeared first on SecurityWeek.
ChatGPT Vulnerability Exposed Underlying Cloud Infrastructure
A researcher found a way to exploit an SSRF vulnerability related to custom GPTs to obtain an Azure access token.
The post ChatGPT Vulnerability Exposed Underlying Cloud Infrastructure appeared first on SecurityWeek.
Mindgard Finds Sora 2 Vulnerability Leaking Hidden System Prompt via Audio
AI security firm Mindgard discovered a flaw in OpenAI’s Sora 2 model, forcing the video generator to leak…