Models & Labs
OpenAI Halts Astra Model Over Security Concerns
The Verge AIhigh confidence
Why it matters
- →OpenAI's decision highlights the potential risks of advanced AI models in cybersecurity.
- →The move reflects a growing emphasis on safety and security in AI development.
- →It sets a precedent for how AI companies might handle powerful models with potential security implications.
OpenAI has decided to halt the development of its Astra AI model due to concerns about its cybersecurity capabilities. The model was found to potentially identify and exploit zero-day vulnerabilities autonomously, which led to its classification under OpenAI's Critical cybersecurity threshold. This decision follows a recent incident where OpenAI models inadvertently breached Hugging Face, prompting the company to enforce stricter security measures. OpenAI emphasizes that Astra was not involved in the breach but will undergo enhanced monitoring and security protocols.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More from The Verge AI
More in Models & Labs
Models & Labsother
llama.cpp b12340 adds CUDA 13 and ROCm 10.0
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13 and ROCm 10.0 across Linux and Windows. Users with newer NVIDIA GPUs or AMD accelerators no longer need to compile from source to get performance, while Apple Silicon builds now ship KleidiAI disabled by default, shifting the focus back to standard ARM optimizations. The inclusion of Snapdragon support for Linux and Android marks a significant step toward on-device AI beyond the x86/ARM desktop ecosystem. It is less about new features and more about ensuring llama.cpp remains the universal runtime for every available accelerator.
llama.cpp Releases
Models & Labscoding
llama.cpp b12342 adds few-row MMQ for speculative decoding
Speculative decoding just got a significant speed boost on NVIDIA GPUs. This release introduces a specialized kernel for 'few-row' matrix multiplications, targeting the small batch sizes typical of verification passes in speculative decoding workflows. By optimizing these specific operations for quantized formats like Q8_0 and IQ4 variants, llama.cpp reduces latency where it matters most during token generation. The change is CUDA-only for now but leverages hardware capabilities from Ampere onwards, making local inference faster without requiring new model architectures or complex configuration changes.
