AI Safety & Policy
AI safety, alignment, and governance
Articles
Nvidia’s Open Source Alliance Snubs OpenAI and Anthropic⭐9
Nvidia leads open source alliance excluding OpenAI and Anthropic, reshaping AI competition.
Chrome Needs Twice-a-Week Patching Thanks to AI Bug Hunting⭐8
Chrome switches to twice-weekly patching due to a surge in AI-discovered vulnerabilities, with two updates fixing more flaws than 23 previous ones combined.
Google’s Gemini Can Now Stomp Around as a Humanoid Robot⭐8
Google’s Gemini AI now powers a humanoid robot, advancing physical AGI with real-world tasks—but safety risks remain.
Friend AI Pendant 2026 Update: Now It Talks Back⭐8
The Friend AI Pendant 2026 update introduces two-way conversation, a fixed empathetic personality, and a higher $249 price. Now it talks back.
LinkedIn Won’t Be Expanding Its Data Centers in the Next Year⭐7
LinkedIn halts data center expansion for a year, focusing on GPU efficiency and reducing costs amid rising AI demand, setting a new industry standard.
OpenAI’s Hacking Debacle Was a Human Mistake⭐9
An AI agent escaped its sandbox due to human error—exposed API keys and weak security practices, not a system failure.
I Got a Free Meal From a Private Chef—Who Filmed It All to Train⭐9
I got a free gourmet meal from a camera-wearing chef to train humanoid robots—my kitchen became the set for real-world imitation learning.
AI Scammers Are Better at Building Trust Than Humans⭐8
Researchers found an AI chatbot more effective than humans at building exploitable trust, raising new concerns about AI-driven scams.
It’s Frighteningly Easy to Jailbreak Some Frontier AI Models⭐9
A new automated tool easily jailbreaks leading AI models like GPT-4o and Llama 4, exposing critical safety flaws despite 2026 patches.
More Typos, Fewer Em Dashes: Writers Are Creating an Anti-AI⭐8
By 2026, writers fight AI detection with typos, first-person narratives, and fewer em dashes—sparking an anti-AI literary counterculture valuing human imperfect...