Full Summary
This Wednesday morning, Microsoft has launched a new agentic security system that **both blockchain.news and Help Net Security confirm** has topped benchmarks. This system, dubbed MDASH, uses over 100 specialized AI agents and has already found 16 vulnerabilities, including four critical remote code execution flaws in Windows, before Patch Tuesday. Microsoft is offering a private preview to customers, highlighting AI's role in digital defense. Meanwhile, Red Hat is also advancing enterprise AI, announcing version 3.4 of its platform. SMBtech reports this update introduces AgentOps for managing autonomous AI agents at scale and a new Model-as-a-Service capability. In a significant development, **vocal.media and The Times of India both report** Anthropic is limiting the release of its new AI system, Claude Mythos, to a small number of organizations, including Microsoft and Google, due to its powerful ability to find and exploit software vulnerabilities. Anthropic believes Mythos is too powerful for public release, sparking debate on AI regulation and alarming US officials after a Chinese representative sought access. On the consumer front, Samsung is launching its 2026 Bespoke AI Laundry Combo, featuring faster wash-to-dry times and fabric-sensing AI, as reported by The Tech Buzz. Meta is expanding its AI Voice Conversation feature, powered by Muse Spark, to India, bringing enhanced natural language processing and new shopping features to WhatsApp, Instagram, and Facebook, according to Deccan Herald. In the business world, FPT AI Factory, InFlow, and Visa Intelligent Commerce have launched an agent-native commerce platform, enabling AI agents to autonomously research, procure, and pay for services, pioneering a "B2AI economy." And in India, AI-powered lending models could unlock a significant credit gap of up to $170 billion for MSMEs, leveraging digital payment trails and GST filings, as noted by english.punjabkesari.com. However, a new study from Pearl Enterprise, highlighted by The AI Journal, reveals that leading AI systems align with expert judgment only about 70% of the time, even though they show gains on public benchmarks. This means companies deploying AI in high-stakes fields like healthcare or finance need to be cautious, as current top systems may not yet match true expert-level performance.