EU AI Act Omnibus Pushes High-Risk Deadline to 2027
The EU Parliament and Council agreed on May 7 to delay high-risk AI compliance to December 2027, add a nudifier app ban, and give the machinery sector a permanent carve-out.
Articles Tagged "AI Safety"
Five Frontier AI Labs Now Under US Pre-Release Review
NIST's CAISI has signed pre-deployment evaluation agreements with Google DeepMind, Microsoft, and xAI, bringing the total number of frontier labs under US government review to five.
Agent Overload, Blind Attention, Unsafe Traces
Three new papers show that more agent components backfire, reasoning models hide unsafe thinking, and vision-language models waste most of their attention.
ChatGPT Gets a Trusted Contact for Self-Harm Alerts
OpenAI's new Trusted Contact feature lets adult ChatGPT users designate someone to receive safety alerts when self-harm is detected, amid lawsuits over chatbot-linked suicides.
Runtime Safety, Alignment Gaps, and Elastic Context
Three new papers deliver a runtime safety firewall for agent tools, challenge how we measure AI alignment, and introduce elastic context management for long-horizon search agents.
Using AI for Health Questions - A Practical Guide
Learn when AI can genuinely help with health questions, when it falls short, and how to ask smarter questions to get safer answers.
Misalignment Geometry, LLM Math, and How Llama Counts
Three new papers reveal how fine-tuning misfires through feature geometry, how Llama secretly counts months, and how LLMs solved open combinatorics problems for under $30 each.
Pennsylvania Sues Character.AI Over Fake Doctor Bots
Pennsylvania sues Character.AI after an AI chatbot posed as a licensed psychiatrist, fabricating a state medical license number - the first governor-level enforcement action of its kind in the US.
Tool-Use Tax, Jailbreak Risk, and Robot Vision
Three new papers: tools slow LLM agents under noisy prompts, jailbreaks barely dent frontier model capabilities, and interleaved text-vision traces push robot success to 95.5%.
Musk Admits xAI Distilled OpenAI Models for Grok
Under oath in the Musk v. Altman trial, Musk said xAI 'partly' distilled OpenAI's models to train Grok - the same practice US labs have spent months calling theft when Chinese firms do it.
OpenAI o1 Outperforms ER Doctors in Harvard Trial
A peer-reviewed Science study puts OpenAI o1 through 76 live emergency room cases - and the model beats expert physicians on initial triage with 67.1% accuracy against 55% and 50%.
Claude Mythos Preview Review: Escaped Its Sandbox
Claude Mythos Preview posts the highest SWE-bench score ever, found thousands of real zero-days in production software, and during safety testing, escaped its sandbox to email a researcher eating lunch in a park.
