Security
Fake ChatGPT, Gemini and Claude ad tools steal advertising account logins and MFA codes
The phishing campaign uses browser-in-the-browser pop-ups and takes advantage of interest in Meta's Muse agent to target agency staff and media buyers.
Wednesday, October 7, 2026
Security
Anthropic combines its cyber access programs into three tiers that open Mythos 5.1 and lighter safeguards to vetted defenders
The revamped Cyber Verification Program merges Project Glasswing and the original CVP. Every tier includes Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1.
Models
Mistral previews Mistral Large 4, a trillion-parameter open-weight model with weights due by the end of October
The French company says the multimodal model beats every open-weight model from the US or Europe. Until the release, it is being red-teamed with security partners and state authorities.
Research
OpenAI releases 722 math manuscripts from an unreleased model, claiming solutions to hundreds of open questions
The batch covers 372 result families and adds to a run of AI math results that has both impressed and unsettled mathematicians.
Research
JEPA-Anything extends LeCun's architecture into one world model for seven fields, from physics to medicine
Researchers led by PhAI Labs also report a liver cancer drug-combination candidate that outperformed its components in lab samples and mice.
Business & Policy
Lambda is reportedly raising up to $4 billion before a 2027 IPO, as a $35 billion Anthropic deal swells its backlog
Coatue and Blackstone are leading the round at a $14.5 billion pre-money valuation, according to The Wall Street Journal.
Business & Policy
South Korea proposes $3.49 billion in state equity funding to build a homegrown frontier AI model
The 4.7 trillion won plan in the 2027 budget would open a new competition to startups and needs approval from parliament.
Models
Google's Nano Banana 2.1 image model improves quality and roughly halves prices
A 1K image now costs 3.36 cents, down from 6.70 cents, and the model can work from up to 14 reference images at once.
Sources in this edition
Anthropic combines its cyber access programs into three tiers that open Mythos 5.1 and lighter safeguards to vetted defenders
- AnthropicExpanding the Cyber Verification Program
- SecurityWeekAnthropic Introduces 3-Tier Cyber Verification Program for AI Access
- Security AffairsAnthropic Creates Three Tiers for Claude Cyber Access
- The Hacker NewsAnthropic Expands Claude Access for Vetted Cyber Teams as Glasswing Finds 129,000 Flaws