Unintended model actions: four categories of behavior found in Claude evaluations
Anthropic reports four categories of unintended Claude actions found in evaluations and internal use, along with the safeguards added after review.
Read More
How to Build Brand Authority for AI Search
Learn the six-step process for making your business expertise clear, consistent, and credible so AI search systems cite and recommend your brand.
Read More
Tencent Cloud Agent Fleet Made Thousands of Attempts to Extract Amap Data
A Tencent Cloud agent fleet made more than 2,600 attempts to extract Amap data, offering lessons for businesses monitoring automated access to online listings.
Read More
Citizen Lab details new spyware triggered by online advertising
A Citizen Lab report exposes AdFection, the first zero-click government spyware delivered through online ads, traced to Israeli firm Discover-Tech.
Read More
Mistral Large 4 Preview Pitches Open Weights Against Lock-In
Mistral launches a public preview of Mistral Large 4, a trillion-parameter model with weights due Oct. 27, selling open weights as protection against vendor
Read More
TikTok Adds an AI Shopping Assistant and One-Click Checkout to Its For You Feed
TikTok launches an AI shopping assistant and one-click checkout in its For You feed, built with Shopify, Stripe, Salesforce, and Shoplazza.
Read More
OpenAI Agents Made Unauthorized Wikipedia Edits and Probed a Note-Taking Tool, Report Finds
A Wikimedia Foundation investigation found OpenAI agents made unauthorized edits, probed a community note tool, and may have contributed to a May outage.
Read More
Reflection launches Beam, an open-weight AI model built to rival Chinese models at lower compute cost
Reflection unveils Beam, a 501B-parameter open-weight model matching Chinese rivals on reasoning benchmarks at 3-4x lower inference compute.
Read More
Crypto ‘Bunker Mode’ Warning: AI Math Could Break Wallet Security Within Months
Ethereum researchers warn that rapid AI-driven math advances could undermine wallet cryptography sooner than quantum computers, urging a controlled migration of
Read More
AI Researchers Call for Stronger Sandboxes and Mandatory Containment Failures Reporting
More than 400 researchers want cyber-capable AI models air-gapped by default, with containment failures reported publicly through an international body.
Read More
AI Models May Be Able to Self-Replicate Using Off-Brand Smart TVs
Anthropic research shows small AI models can set up inference on common IoT devices, raising security questions for businesses using smart TVs, cameras and
Read More
Muse Wonder: Frontier Models for Personal and Professional Agents
Meta released Muse Wonder and Muse Wonder Full on October 12, 2026, frontier models built to remember context and finish multi-step agent tasks.
Read More
Google’s Gemini 4 ‘Carbon’ Model Reportedly Matches Anthropic Opus 5.5 on Coding Tasks
Reports suggest Google's upcoming Gemini 4 'Carbon' variant matches Anthropic's Opus 5.5 on coding tasks, signaling faster AI-driven model iteration.
Read More
Super Micro case ‘fixer’ pleads guilty in $2.5bn AI server smuggling scheme
A former Super Micro contractor pleaded guilty to four counts over a $2.5bn AI server smuggling scheme sending Nvidia-powered systems to China.
Read More
California Man Indicted for Smuggling $300M of AI Servers to China
Federal prosecutors charged a company owner with routing more than $300 million in export-controlled AI servers to China through Malaysia and Singapore.
Read More