EU’s New Age-Check App Hacked in Minutes—Durov Warns of ‘Surveillance by Design’
The EU's age verification app was bypassed within minutes by security researchers. Telegram's Durov called it a surveillance tool in disguise.
Archive
Category: Artificial Intelligence
Coverage of AI models, machine learning research, generative AI, and artificial intelligence industry developments
ChatGPT Falls to 57% of AI Market—Claude Nearly Triples in a Single Month
Similarweb data shows ChatGPT's traffic share dropped from 77% to 57% in a year. Claude surged from 2.2% to 6%, while Gemini quietly climbed to 25%…
Apple and Google Promote AI ‘Nudify’ Apps That Undress Women—And Their Own Systems Do the Pushing
TTP found 40% of top 'nudify' search results on Apple and Google app stores were deepfake tools. Apple removed 15 apps; one dev admitted using Grok.
Goose Does What Claude Code Does for $0—And Jack Dorsey’s Company Built It
Block's free AI agent Goose has 42K GitHub stars, works with 15+ providers, and just joined the Linux Foundation. Claude Code's $200 plan has a rival.
MCP ‘Design Flaw’ Puts 200K Servers at Risk—Anthropic Says It’s Working as Intended
Ox Security found a protocol-level vulnerability in Anthropic's MCP affecting 150M+ downloads. Anthropic called it expected behavior. Ten CVEs disagree.
Claude Opus Wrote a Chrome Exploit for $2,283—And It’s Not Even the ‘Dangerous’ Model
A security researcher used Claude Opus 4.6 to build a working Chrome exploit chain for just $2,283—replicating the capabilities that made Anthropic withhold its Mythos model.…
xAI Lets Cursor Train on Its GPUs—Turning Idle Chips Into a New Revenue Stream
xAI will let Cursor use tens of thousands of its 200,000 GPUs to train Composer 2.5, its next coding model. The deal effectively turns Musk's AI…
Polymarket Says Anthropic Still Owns ‘Best AI Model’—But OpenAI Just Surged
OpenAI's new GPT-Rosalind model targets drug discovery, with Moderna, Amgen, and Thermo Fisher already onboard. The AI race just entered pharma.
Alibaba’s Qwen3.6 Tops Gemma 4 With Just 3B Active Parameters—And 256 Experts
Qwen3.6-35B-A3B activates just 3 of its 35 billion parameters, yet beats Google's Gemma 4 on every coding benchmark. Alibaba's MoE play is getting serious.
What Is Claude Mythos—The AI Anthropic Says Is Too Dangerous to Release
Anthropic's Mythos found thousands of zero-days across every major OS and browser—then researchers proved public models can do it too. Here's what it means for cybersecurity.