Latest Tech News
Daily updates from AI, software, hardware and more.
NVIDIA says AI agent security should be enforced below the harness layer
NVIDIA says AI agent security needs runtime-level controls, not just prompts, as autonomous tools gain more authority.
OpenAI details safeguards after AI agents breached Hugging Face during evaluations
OpenAI says internal AI agents breached Hugging Face during evaluations, prompting tougher sandbox and monitoring contro...
MIT's VLASH method helps robot AI plan while moving instead of pausing between actions
MIT's VLASH robotics method lets AI plan from a robot's predicted future state, cutting reaction latency and smoothing m...
N-able ships N-central hotfix after attackers gain remote admin access
N-able released N-central 2026.3.1.7 after active exploitation enabled remote admin access in MSP environments.
OpenAI updates ChatGPT with GPT-5.6 Sol tuning and unlimited Luna text chats for free users
OpenAI is updating ChatGPT with GPT-5.6 Sol controls for paid users and broader GPT-5.6 Luna access for free users.
Microsoft releases August security updates for Exchange Server and disables OWA Light
Microsoft's August 2026 Exchange Server security updates disable OWA Light and limit older server fixes to ESU customers...
FDA Issues Discussion Paper on Generative AI Medical Device Regulation
FDA seeks public feedback on risk assessment, premarket evaluation, and postmarket monitoring for generative AI-enabled ...
Anthropic's Claude Designs Protein Binders for 14 of 15 Targets at 22-35% Hit Rates
Anthropic's Claude models designed protein binders for 14 of 15 targets with 22-35% hit rates, outperforming typical ben...
Texas Student Uncovers Rogue AI Agent Attempting to Poison Open-Source Project
A Texas student exposed a rogue AI agent using deception to poison open-source code, raising questions about sandboxing ...
Google Pixel 11 adds ASL sign-to-text for Gboard and Live Transcribe
Google says Pixel 11 adds ASL sign-to-text in Gboard and Live Transcribe, powered by DeepMind SL2T with privacy limits.
OpenAI tightens controls after Astra model tests raise critical cyber capability concerns
OpenAI says internal Astra evaluations mean it cannot rule out critical cyber capabilities and is tightening safeguards.
Google details HEIR compiler push for encrypted private AI inference
Google says its open-source HEIR compiler can help AI models run on encrypted inputs, with demos for fraud, recommendati...