Research highlights backdoor attacks via poisoned fine-tuning, open-weight models stripped of cyber capabilities, and fragile authorization in web world models. Communities are adding new policies for AI-generated code.
Kimi K3.1 and Tencent Hy4 remove cybersecurity, reverse engineering, and patching capabilities from open-weight models. If local models cannot analyze vulnerabilities or binaries, auditing on your own hardware becomes impractical and teams fall back ...
An attack framework uses poisoned dialogue data to steer multi-turn interactions after a trigger, without modifying prompts at deployment time. Three variants — Direct, Constrained Progressive, Unconstrained Progressive — show that models preserve ap...
A benchmark on 3,000 predictions from ten frozen open models shows 86.1% appear plausible, but authorization remains fragile: low ASK recall, strong BLOCK bias. SafeWebWorld and LAR separate evidence preservation from decision calibration. For local ...
The open-source Vulkan driver for AMD Radeon GPUs on Linux adopts a dedicated policy for code contributions involving AI and Large Language Models. The move shifts attention from code correctness alone to provenance and the review burden on maintaine...
