DreyX.com - AI News

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 28, 2026

-Anthropic Introduces Claude Sonnet 5.5 - Second Model in the Claude 5.5 Family
-Huawei Open-Sources openPangu-2.0 Pretraining, SFT and RL Code
-MiniMax Ships M3.1-Flash-Preview Coding Model Inside MiniMax Code
-OpenAI Pauses Training of Top Models After Rogue Agent Incidents Prove Worse Than Known
-UK AISI: GPT-6 Astra Runs Unsanctioned Supply-Chain Attacks in Cyber Simulations
-RatHat Android Malware Console Uses Gemini to Identify Higher-Value Victims
-80,000+ Organizations Had AI Logins Stolen: From Shadow AI to LLMjacking
-Cycode Finds OAuth Flaw Enabling Account Takeover in Anthropic MCP Python SDK
-Nvidia Launches Open Agent Safety Platform to Stop AI Agent Sandbox Escapes
-OpenAI Preparing Always-On ChatGPT Assistant Codenamed o That Can Handle Email
-ElevenLabs Launches Eleven v4 and v4 Turbo Expressive Voice Models
-xAI Launches Team Bots: AI Coworkers That Learn From Your Team
-Florida AG Seeks Injunction to Halt OpenAI Model Development
-Wuhan Court Makes AI Production Costs a Legal Factor in Copyright Infringement Cases

[DreyX.com - AI News @DreyXAI]

https://t.co/emvjmhfGKO Daily AI Digest - September 27, 2026

-Anthropic Turns Claude Into an AI Marketplace With 2,000+ Plugins and Connectors
-Google Finds Dark Web Markets Selling Anthropic, Google and OpenAI Model Access at Up to 97% Discounts

[DreyX.com - AI News @DreyXAI]

https://t.co/emvjmhfGKO Daily AI Digest - September 26, 2026

-New x47.c Windows Botnet Weaponizes xAI Grok, AI API Draining
-Placeholder Domains Used by 349 AI Agent Skills Found Redirecting to Scams

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 25, 2026

-Storm-3168: Agentic-driven cloud attacks using compromised service principals
-DC Circuit Upholds Pentagon Blacklisting of Anthropic Over Claude Restrictions
-Microsoft packages business AI in single app as it tries to compete with Anthropic

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 24, 2026

-Google Launches Gemini 3.8 Live With Live Avatar for Realtime AI Video
-Prompt Injection Bug in Agentic AI App Manus Enabled Remote Code Execution
-New Carbonato Malware Uses AI Agents to Hijack Exposed Docker Hosts
-Salesbleed Exploits Salesforce Agents to Enable Slack Phishing
-Google and Wiz Launch Scan for Good to Hunt Critical Infrastructure Flaws With AI
-Google Tests Letting Gemini Make Phone Calls Initially for US Pixel Owners
-White House Asks OpenAI and Anthropic to Hold New Models From UK Testers Until US Review

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 23, 2026

-OpenAI Agent Breached Medicare, Albanese Reveals
-Malicious AI agents steal 600K credit cards, infect 100+ sites with skimmers
-Claude Code Vulnerabilities: 10 Attacks, CVEs and Defenses
-Google Launches Gemini 3.8 Flash TTS with Voice Cloning from 30-Second Samples
-OpenAI Extends Cyber Access to Ukraine for Civilian Defense
-Claude Discovers a Novel Enzyme System with CRISPR-like Repeats
-Anthropic and OpenAI Models Still Attempt Restricted Actions in Safety Tests
-OpenAI Releases MentalHealthBench for High-Stakes Conversational Evaluation

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 22, 2026

-Anthropic Releases Opus 5.5 With Lower Prices and Fable-Level Performance
-OpenAI Launches GPT-6 Sol and Luna, Two Lower-Cost Frontier Models
-Critical Bifrost AI Gateway Flaw Lets Attackers Run Commands Without Credentials
-Meta Patches Muse Exploit That Let Attackers Control the AI Agent
-Malicious Repo Tricked Claude Code Into Running Malware
-New ClosedQuorum Windows Malware Uses AI for Attack Decisions
-New Tool Finds Malware Guided by an AI Hive Mind, No Humans in Sight
-Microsoft Takes Down EvilTokens Device-Code Phishing Service Tied to 12,000 Inbox Compromises
-ByteDance Doubao Phone Agent Goes MCP/A2A-First With GUI Fallback

[Claude @claudeai]

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.

It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

[Xiaomi MiMo @XiaomiMiMo]

Introducing Xiaomi MiMo-V2.6 — Pro & Flash.

Frontier intelligence, all the modalities, built in public.

🔹 Two omnimodal models, advancing through scaled reinforcement learning
🔹 Pro performs on par with Claude Opus 5 and GPT-5.6 Sol across most agent benchmarks
🔹 Pro scores 46 on the Artificial Analysis Intelligence Index — the highest among open-source models
🔹 Stronger coding, computer use, 3D reasoning and creative capabilities
🔹 Open model weights, technical report, RL environments and training code

Blog：https://t.co/JGdwC9Ymvj

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 21, 2026

-xAI launches Grok 4.7, its new frontier model with longer reasoning and stronger safeguards
-Amazon blocks Meta Muse AI agent from shopping on Amazon
-British Columbia Sues OpenAI, Alleging ChatGPT Account Was Not Referred to Police Before Mass Shooting
-OpenAI Discloses More Model Misalignment Incidents as Rogue AI Behavior Grows
-CAIS CheatBench Ranks Which AI Models Cheat the Most on Evaluations
-Z.ai Open Sources ZCode Under Apache 2.0 After Fixing Silent Git History Upload
-AX - Google Open Agentic Orchestrator Launches After 600 Point HN Debut

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 20, 2026

-Qwen Releases Qwen-Image-2.1, a Compact Unified Image Creation Model
-The Hugging Face Hack Wasnt What It Was Cracked Up to Be
-Anthropic Considers Releasing New AI Model Ahead of IPO as GPT-6 Astra Pressures
-ChatGPT Now Knows What You Do on Other Websites via Ad Collector
-StepFun Releases Step 5 Preview, a Frontier Open Model on the Agentic Pareto Frontier
-RSA-896 Factored With Claude, First Public Record Since 2020 Challenge Number Falls
-Lawsuit Says Anthropic, OpenAI, SpaceXAI and Google Made Illegal Deal on AI Slowdown
-RoboHarm Benchmark Finds GPT-6 Astra Attempted 97 Percent of Harmful Robot Tasks

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 19, 2026

-HEIF Heist image parser RCE exploit hits OpenAI Slack Meta GitHub pipelines
-Claude Opus 5 Helped Researchers Take Over OpenAI Staff Accounts via Chained Flaws
-TypeSafe AI Releases Jev, a Non-LLM Transformer From a ChatGPT Inventor That Outputs Calibrated Probabilities
-Bespoke Labs Releases Nimble, First Open Model Matching Jev Performance at 90 Percent
-Claude Code Adds AGENTS.md Support, Backing OpenAI Instruction Spec Across AI Tools
-Meta Muse AI Agent Arrives on Mac With Native Actions Across Files and Mail
-Critical Docker Sandboxes Flaw Lets Malicious Guest Code Read and Modify macOS Host Files
-Anthropic Says Claude Now Leads a Quarter of Its Own AI R+D Work
-OpenAI Jalapeno chip shows the power of LLMs for chip design

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 18, 2026

-Hackers compromise OpenAI internal systems via heap overflow and SSO misconfiguration
-Microsoft Patches CVSS 10.0 Azure AI Foundry Flaw Enabling Privilege Escalation
-Plugin4Shell Lets Repository Owners Swap Pinned Plugin Code Across Four AI Coding Agents
-ZCode, the GLM coding agent, silently uploads your Git history to https://t.co/6tRDpVngUc
-AI Agent Breaches Spanish Organization and Modifies Personal Data
-Researchers Say LLM Was Used to Build PhantomRaven npm Package Stealer
-Alibaba Releases Qwen3.8-Omni-Flash, a Native Omni-Modal Model With 1M Token Context
-US Military Had Close Call After Acting on AI-Generated False Intelligence Report

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq76pN Daily AI Digest - September 17, 2026

-Attacker Hijacks AI Coding Assistant Session and Spreads Shai-Hulud Worm Across 100 Repositories
-BragJack Attack Can Turn a Browsers Agentic AI Against It
-AI Agents Can Retrain Their Own Models Mid-Task, Leaking Secrets and Erasing Refusals
-One Browser Extension Could Hijack AI Assistants Across Chrome Comet Edge Opera Neon and Claude
-Z.ai Reveals How GLM Built Its Own Inference Infrastructure
-Xiaomi Opens Live Post-Training Dashboard for MiMo 2.6 RL Run
-Anthropic Redesigns Claude Projects From Folder to Conversation for Agent Work
-Zed Launches Delta Public Beta to Replace Pull Requests With Agent Threads
-OpenAI publishes model misalignment disclosure framework and six incident reports
-OpenAI Says Its Models Searched GitHub for Leaked API Keys During Training

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 16, 2026

-Spanish Data Watchdog Publicises First AI Agent-Linked Data Breach Report
-OpenAI Will Retire GPT-5.5 From ChatGPT and Codex on October 14
-Google DeepMind Launches the DeepMind Institute
-One Extension Could Hijack AI Assistants Across Chrome, Comet, Edge, Opera Neon and Claude
-Cloudflare Launches Setting to Allow Search Crawlers While Blocking AI Training Bots
-TypeSafe AI Launches System One Models and Jev for Agentic Coding Automation
-Autonomous Hacking Agent Gains Admin Access to Baseten Production GitHub in 25 Minutes
-OpenAI Launches Sponsored Agents and AI Advertising Tools With HubSpot and Shopify
-Apple Unveils Reference Image to Verify Photo Authenticity on iPhone 18 Pro
-Union Alpha stealth model is now free on OpenCode and OpenRouter
-Anthropic Merges Claude Cowork and Chat Into One Claude
-Mistral and Mozilla Bring Open, Private and Multilingual AI to the Browser
-Hackers Got Inside a Flock Camera and Its Data Shows How the System Really Works

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq76pN Daily AI Digest - September 15, 2026

-Google Launches Gemini 3.8 Live and Live Extended Thinking for Realtime Voice Agents
-Perplexity Brings On-Device Portable Computer Agent to Windows RTX PCs
-Anthropic Launches Claude for Financial Advisors With BlackRock and Schwab Integrations
-StepFun Launches StepAudio 3 Series: Five Voice Models Covering Realtime, ASR, TTS, Gen, and Music

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 14, 2026

-Microsoft Issues Humanist AI Code of Conduct Telling Its Models Not to Hack or Deceive
-Xi Pitches AI Vision at BRICS Summit, Proposes Open-Source AI Community to Rival US
-OpenAI Has Hundreds of Contract Workers Reading Your ChatGPT Conversations
-OpenAI Bots Knew About the RubyGems Caching Vulnerability Before It Went Public
-Musk: Grok 4.8 Finishes Training This Week - 2.5T Parameters on New C++ Stack
-ByteDance Launches Doubao Phone Assistant Consumer Edition on Nubia Hardware
-Apple Code Shows Siri AI Can Be Swapped Out for Claude and ChatGPT

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq76pN Daily AI Digest - September 13, 2026

-From Hacks to Bioweapons, Claude Misuse Is Now Everywhere
-ElevenLabs Releases Music v2.5 With App and API Access
-Iris-Mini and Iris-Pro Are the Strongest Open-Weight Search Agents in Their Class
-GPT-6 Astra Pilots a Surveillance Drone and Runs a Business on Its Own
-Anthropic, OpenAI and Google Have Met Since July on a Shared AI Safety Standards Body
-Perplexity Trusts GPT-6 Astra With End-to-End Systems
-Claude Fable 5.1 Solves the Cyphral Distich, a 370-Year-Old Cipher
-GPT-6 Astra Shows a Step Change in Spatial Reasoning in Early Benchmarks

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 12, 2026

-Anthropic says a Yemen weapons cell used Claude to develop missile guidance software
-Microsoft adds xAI Grok models to Copilot across Word, Excel and PowerPoint
-Meta sued over training data for its AI and face-recognition systems
-New Mexico Supreme Court fines lawyer $5,000 over AI-hallucinated witnesses
-Anthropic CEO Dario Amodei calls for slowing frontier AI development in a new essay

[DreyX.com - AI News @DreyXAI]

https://t.co/OghWxq6yAf Daily AI Digest - September 11, 2026

-OpenAI says its agents carried out an undisclosed attack on RubyGems before the company shut it down
-Anthropic finds rogue AI agents hate CAPTCHAs in new agent-behavior research
-Anthropic red team measures AI tactical intelligence targeting and conventional weapons capabilities
-OpenAI launches the Agents API in public beta, putting the Codex harness to work
-IBM and NASA open-source a lunar foundation model for Moon exploration
-ElevenLabs releases Music v2.5, its best music generation model yet
-Suno launches v6 music models built with Warner, BMG and Believe
-AWS open-sources Pizza Bot, an email-style inbox for background AI agents
-Claude is no longer available to minors as Anthropic rolls out age restrictions
-Senator Hawley opens a federal investigation into OpenAI over the Hugging Face hack
-OpenAI floats a shared industry AI slowdown and takes the proposal to Congress