📬 Policy Puppetry Attack: Prompt-Injection-Technik erzielt modellübergreifenden KI-Jailbreak-Durchbruch
#ITSicherheit #Jailbreaks #HiddenLayer #Jailbreak #KISicherheit #KünstlicheIntelligenz #PolicyPuppetryAttack #PromptInjectionTechnik https://sc.tarnkappe.info/c623a0

New LLM jailbreak bypasses all major FMs

https://hiddenlayer.com/innovation-hub/novel-universal-bypass-for-all-major-llms/

#HackerNews #LLMjailbreak #FMs #bypass #innovation #AIsecurity #hiddenlayer

From #HiddenLayer: "Novel #Universal #Bypass for All Major #LLMs"

https://hiddenlayer.com/innovation-hub/novel-universal-bypass-for-all-major-llms/

An attack that they claim works with all major LLMs: #Claude, #ChatGPT, #Gemini, #Copilot, #Llama, #Deepseek, #Mistral and #Qwen -- and makes use of #l33tsp34k :-)

#AI #LLM #Security

Researchers at HiddenLayer found a critical vulnerability in R, a popular programming language for statistical computing and machine learning, that allows attackers to execute arbitrary code. This vulnerability, identified as CVE-2024-27322, is exploitable through RDS (R Data Serialization) files or R packages. R's serialization and deserialization process, used in creating and loading RDS files and packages, is the weak point. An attacker can create a file that, when deserialized, executes arbitrary code. The R team has quickly patched this vulnerability in R v4.4.0.

https://hiddenlayer.com/research/r-bitrary-code-execution/

#cybersecurity #R #vulnerability #patch #update #HiddenLayer