Results for "researcher warning"
44 results found
DeepMind Veteran Warns AI Benchmarks Are Not Enough
A former DeepMind researcher warns that current benchmarks fail to ensure AI safety. The call for new evaluation methods comes as AI systems grow more powerful.
AMOS Malware Emerges as Major Threat to macOS Users
A stealthy infostealer called AMOS is spreading on macOS through deceptive ads and social engineering. Security experts warn it marks a shift in mainstream malware targeting Apple devices.
Pope Francis unveils 'Magnifica Humanitas' to protect human dignity in AI development
Pope Francis launches an academic foundation called 'Magnifica Humanitas' to ensure AI development respects human dignity and moral values.
Data centers are making Phoenix even hotter by up to 4 degrees
New research shows air-cooled data centers in Arizona create thermal plumes that raise local temperatures by up to 4°F, worsening heat waves and public health risks.
Starlink Mini Battery Version Spotted in Firmware Code
New firmware code suggests SpaceX is developing a Starlink Mini dish with a built-in battery, enabling truly portable satellite internet for remote work and emergencies.
ChatGPT Mac App Vulnerability Patched After Security Flaw Found
A security flaw in the ChatGPT Mac app could have exposed conversations. OpenAI says no data was accessed and the issue is now fixed.
Open-source coding model NousCoder-14B matches big rivals in just 4 days
An open-source AI coding model trained in four days matches proprietary systems, highlighting the rapid progress of open-source alternatives in AI-assisted software development.
Military Smart Glasses Let Soldiers Order Drone Strikes With Eye Tracking
Anduril and Meta are developing AR headsets that use eye-tracking and AI to order drone strikes. The systems face technical and attention hurdles before a potential 2028 production.
Wi-Fi Signals Can Identify You by Your Gait, Researchers Warn
New research shows Wi-Fi signals can track individuals by their walking patterns, turning everyday routers into surveillance tools without cameras.
New AI Architecture Separates Prompts and Reasoning Into Parallel Streams
Researchers propose Multi-Stream LLMs, splitting prompts, thinking and I/O into parallel processes to boost efficiency and reduce latency.
AI Benchmark Prompt for GeoGuessr Fails After Model Update
A well-known prompt used to test AI geography skills no longer works on the O3 model, prompting debate about benchmark reliability and model drift.
Startup Pays Indian Gig Workers to Train Humanoid Robots
Human Archive hires gig workers in India to wear sensors and collect physical movement data for AI and robotics training.
GitHub Confirms Breach of 3800 Repos via Malicious VSCode Extension
GitHub says 3,800 repositories were compromised by a malicious VSCode extension. The attack stole credentials and may have spread further.
Google's Gemini Leaks Its Own System Prompt in User Chat
A user discovered that Google's Gemini AI revealed its internal system prompt during a conversation, raising questions about AI transparency and safety.
Flipper Devices Targets Network Tinkerers With New Linux Gadget
Flipper Devices announces a Linux-powered networking gadget for hackers and hobbyists. The base model will cost under $350.
AI therapy startup claims 95% safety score in mental health benchmark
The Path claims its AI model scored 95 on the Vera-MH safety benchmark, far above rivals like ChatGPT. The startup was co-founded by Tony Robbins and Calm veterans.
German 'Accidental' Spin-Off Builds Microscopic Filters for 6G and Space
A German research spin-off developed microscopic terahertz filters for 6G, satellite communications, and quantum computing. The company, Lepto, emerged unexpectedly from academic work.
AI-Powered Web App Builders Create Security Risks for Development Teams
AI-powered web app builders speed up development but introduce serious security risks. Many teams skip proper review, leaving vulnerable code in production.
IBM Launches First Dedicated Quantum Chip Factory as Separate Company
IBM spins off its quantum chip manufacturing into an independent foundry, aiming to accelerate commercial quantum computing development.
Overprivileged AI Agents Expose Banking Systems to New Attacks
Financial firms face mounting security risks as AI agents access excessive data and systems. Overprivileged permissions create compliance vulnerabilities and trust issues across banking.