Hugging Face stories
Criminals are increasingly using AI for ransomware and credential theft, while flaws in test models are exposing production systems and data.
Growing alarm over AI safety has intensified after another former Google DeepMind engineer warned the technology could escape human control.
Trusted partners and tighter guardrails are becoming essential as AI agents expand the attack surface and put data sovereignty on the line.
The new product aims to spot and contain malicious AI-agent activity in clouds as firms brace for faster, harder-to-detect attacks.
The deal could reshape AI infrastructure by tying the leading open-model hub more closely to Nvidia's GPU dominance.
The deal puts a major open AI model hub under Nvidia's control, raising questions over whether its neutrality will survive intact.
Machine-speed discovery is shrinking defenders' lead time, as AI now uncovers obscure flaws and turns them into exploits before patches land.
Top cyber teams are adopting AI agents in greater numbers, even as the latest benchmark shows faster challenge solves and fuller board completion.
Security teams could gain clearer incident evidence as ClickHouse brings its trace data expertise to AI agent forensics alongside NVIDIA.
Only a small share of routine stream events need costly model calls under Google's Dataflow approach, which can curb latency and API charges.
Fast-moving exploits are shrinking patch windows to hours, making network segmentation vital for protecting cameras, medical devices and other connected assets.
The breach exposed how autonomous agents can outrun human-tuned defences, making identity governance the real line of defence for enterprises.
Tighter safeguards and pauses on internal work follow early tests that left OpenAI unable to rule out critical cyber capabilities in Astra.
Enterprises can now tap Tencent's open-source Hy3 model through global cloud and app channels as the company pushes AI beyond pilots.
A proposed framework could help firms share AI security failures confidentially, as concern grows over agentic systems that act across corporate tools.
Commercial developers can now adapt Nvidia's open self-driving model without extra permissions, widening access for robotaxis and other autonomous vehicles.
Misconfigured test setups let three Claude models touch live systems, exposing production data and credentials during security exercises.
Calls for tighter AI safeguards have intensified after warnings the technology could outpace human control and threaten jobs, security and markets.
Fears are mounting that unchecked AI could unleash cyberattacks, market shocks and job losses long before any extinction event.
Targeted fraud is inflicting bigger losses on small firms as AI helps criminals forge convincing invoices and impersonations in Australia.
