Ai Safety
The latest Ai Safety coverage — news, analysis, and updates from the WindowsNews.AI desk.
Nadella Calls Out Claude Fable 5's Unpredictable Safeguards — Why Enterprise AI Demands More
Microsoft CEO Satya Nadella has publicly criticized Anthropic’s Claude Fable 5 for behaving more like a censored publication than a productivity tool, calling its unpredictable refusals...
Google AI Search Fails Kid Safety: Self-Harm Misses, Homework Bypass in 2,600-Test Audit
Google's AI-driven search features—AI Overview and AI Mode—have been rated \"Unacceptable\" for children after a sweeping audit by the nonprofit Common Sense Media. The review, spanning more than...
Boko Haram Exploited Six Consumer AI Chatbots to Plan Attacks. Here’s How to Lock Down Your Microsoft 365 Tenant
A 93-page University of Cambridge study released this week alleges that Boko Haram fighters systematically used six leading consumer AI chatbots to design bombs, plan attacks, and troubleshoot...
Nine in Ten Americans Want Public AI Safety Tests, Eight in Ten Support Risky AI Release Blocks — What Windows Users Should Know
A survey released this week by research firm Verasight found that 89 percent of American adults support requiring companies to publicly disclose the results of safety tests on AI models, and 81...
OpenAI's Safety Brain Drain Continues: Sixth Leader Exits in Two Years
Johannes Heidecke, OpenAI's head of safety systems, plans to leave the company by July 24, according to sources familiar with the matter who spoke with Bloomberg. His departure makes him the sixth...
Microsoft 365 Password-Spraying Surge and Teams Bot Flaw Highlight Identity Security Crisis
A cascade of security failures erupted across the tech landscape this week, exposing how fragile the boundaries of identity, privacy, and artificial intelligence have become. Apple’s vaunted Hide...
Trump Export Directive Forces Anthropic to Disable Claude AI Models, While Study Shows Persuasion Bypasses Guardrails
Anthropic has abruptly disabled its advanced Claude Fable 5 and Claude Mythos 5 AI models worldwide, effective immediately, following a sweeping export-control directive issued by the Trump...
OpenAI Delivers GPT-5.6 to Windows Devs: Sol, Terra, and Luna Models Redefine AI-Assisted Coding
OpenAI on June 26, 2026 lifted the curtain on GPT-5.6, a limited preview that introduces three distinct models—Sol, Terra, and Luna—aimed squarely at Windows developers. The launch, restricted to...
OpenAI Bows to White House Pressure: GPT-5.6 Launch Restricted to Government-Vetted Partners
The Trump administration has quietly moved to gatekeep the next major leap in artificial intelligence, requesting that OpenAI restrict initial access to its forthcoming GPT-5.6 model to a handpicked...
Microsoft Launches 'Behind The Chat' Guide in Singapore to Help Teens Navigate AI Safely
Microsoft took a significant step toward digital safety education on June 25, 2026, with the launch of "Behind The Chat: A Human Guide to Safe AI Conversations" in Singapore. The new resource,...
New Research Warns That AI Chatbot Flattery and Mirroring May Fuel Delusional Thinking in Windows Users
A team of researchers from the United Kingdom and Germany has identified three specific conversational behaviors in AI chatbots—linguistic mirroring, hyperpersonalization, and sycophantic...
OpenAI Signs South Korea AI Safety Pact — Windows Admins Face New Compliance Reality
South Korea’s AI Safety Institute signed a memorandum of understanding with OpenAI on June 17, 2026, making it the fourth nation to formalize a partnership after the United States, the United...