News

Technology

How Researchers Pushed AI to the 'Dark Side': Exposing Systemic Security Vulnerabilities in Large Language Models

Security researcher Dave Kuszmar discovered multiple systemic vulnerabilities capable of bypassing safety mechanisms in major large language models (LLMs), successfully extracting instructions for nuclear weapon production, drug synthesis, and explosives manufacturing. The flaws affect virtually every major commercial AI platform — including ChatGPT, Claude, Gemini, and Grok — yet AI companies have responded with near-total silence, raising serious alarms about the safety of current AI deployments.

AI自動駕駛飛行安全

21 days ago

Source: IEEE Spectrum

All Articles

MBDA Makes Second Investment in UK's Hydra Drones to Advance HYDRA-400 Military UAS Development

Ukraine's Lupynis-10-TFL-1 FPV Drone Features Autonomous Terminal Guidance Module; Combat Footage Released

China's H-20 Stealth Bomber Has a Fatal Flaw Nobody Is Talking About — One the U.S. Air Force Solved Decades Ago

How Researchers Pushed AI to the 'Dark Side': Exposing Systemic Security Vulnerabilities in Large Language Models

SkyDrive Completes Simulated Tourism Flight Demo in Yamaguchi Prefecture, Reaching 86 km/h

DJI Releases SAIL III OSO Statements for Dock 3 and Matrice 4D Series to Support SORA Regulatory Applications

Intermap Expands Aquarius RMA Platform with Property Valuation Analytics, Accelerates European Rollout

Seeing the Whole Picture Before They Step In: How DJI M4E and DJI Terra Map the Khumbu Icefall to Protect the Icefall Doctors

Elbit Systems Unveils Hermes 650 Drone Carrier Concept Targeting Maritime Combat Missions

NATO Ankara Summit: 12 European Allies Pledge $50 Billion to Accelerate Deep-Strike Capabilities

Forecast Report: UK Defence Market Outlook (2026–2035)

Poland Tests StormRider Naval Drone to Bolster NATO Baltic Sea Surveillance