Microsoft has unveiled MDASH, a new AI system designed to discover and validate software security flaws at scale. The platform utilizes over 100 specialized AI agents to scan code and automatically verify findings. In testing against a private Windows driver containing 21 intentional flaws, the tool successfully identified every vulnerability without generating any false alarms. It also achieved a 96% success rate on historical Microsoft security data and topped the public CyberGym benchmark.
Microsoft patched the 16 previously unknown security flaws MDASH discovered in Windows during its May 12 enterprise update, including four critical remote code execution vulnerabilities. The system is currently available in a limited private preview for enterprise customers and is model-agnostic, allowing Microsoft to swap underlying AI models without altering the main architecture.
Meanwhile, Fortinet has updated its FortiAIGate platform with security features powered by NVIDIA Blackwell and Hopper GPUs. This solution provides zero-trust security for autonomous AI agents and large language models, helping prevent prompt injection attacks and unauthorized content generation. IBM has also expanded its cloud offerings with Red Hat AI Inference and Red Hat OpenShift Virtualization Service, enabling enterprises to run production AI models and manage virtual machines more efficiently.
Outside of enterprise security, AI adoption is accelerating across various sectors. Booking.com CEO Glenn Fogel noted that AI is fundamentally changing how the travel and airline industry handles bookings and customer interactions. In healthcare, Databricks released the Site Feasibility Workbench to unify data sources and predict clinical trial site success, while a UK study found that 14 percent of people prefer AI chatbots for medical advice due to convenience and cost.
Broader industry shifts include data centers in the US and UK now consuming 6 percent of national electricity supply, a figure driven by growing AI demand. Penn State Extension is hosting a free webinar on May 15 to help farmers apply AI tools to marketing. Additionally, OpenAI held a Parameter Golf competition where over 1,000 participants used AI agents to optimize code under strict constraints, highlighting new techniques in model compression.
Key Takeaways
["Microsoft AI Tool Finds 16 Windows Security Flaws
Microsoft released a new AI system called MDASH that discovered 16 previously unknown security flaws in Windows. The tool identified four critical remote code execution vulnerabilities and 12 other important issues. Microsoft patched these problems in its May 12 update for enterprise customers. The MDASH system uses over 100 specialized AI agents to scan code and validate findings automatically. This technology marks a shift toward using AI to find and fix software security problems faster than traditional methods.
Microsoft AI System Validates Security Flaw Detection
Microsoft tested its new MDASH AI security system against a private Windows driver with 21 intentional flaws. The AI tool successfully found all 21 vulnerabilities without any false alarms. The system achieved a 96% success rate on historical Microsoft security data and topped a public benchmark called CyberGym. Microsoft is currently offering the tool in a limited private preview for enterprise customers. The results show that AI agents can match the performance of professional security researchers.
Microsoft Launches MDASH AI for Vulnerability Discovery
Microsoft unveiled MDASH, a new AI system designed to find and fix software security flaws at scale. The platform uses multiple AI agents to scan code, validate findings, and prove that vulnerabilities are exploitable. It is currently available in a limited private preview for enterprise customers. The system is model-agnostic, meaning Microsoft can swap out the underlying AI models without changing the main architecture. This tool helps companies respond faster to the growing threat of AI-powered cyberattacks.
Fortinet Adds NVIDIA AI Security to FortiAIGate
Fortinet updated its FortiAIGate platform with new security features powered by NVIDIA hardware. The solution uses NVIDIA Blackwell and Hopper GPUs to protect AI workloads in real time. It provides zero-trust security for autonomous AI agents and large language models. The system can prevent prompt injection attacks and stop unauthorized AI-generated content. Companies can deploy it on-premises or in the cloud to keep their AI applications safe and compliant with privacy laws.
IBM Adds Red Hat AI and Virtualization Services
IBM announced two new managed services on its cloud platform built with Red Hat technology. The first, Red Hat AI Inference, helps companies run production AI models without managing complex hardware. The second, Red Hat OpenShift Virtualization Service, allows businesses to move virtual machines to a modern container-based environment. These tools are designed to help enterprises move from AI experiments to real-world applications. They offer consistent performance and predictable costs for scaling AI workloads.
Data Centers Use 6 Percent of US and UK Power
New research shows data centers in the UK and US now use 6 percent of the national electricity supply. This figure has risen 15 percent globally in the last two years due to growing demand for AI. The International Data Center Association warns that this energy use is causing community resistance and political pushback. Some countries like Singapore and Lithuania see even higher usage at 19 percent and 11 percent respectively. Experts call for better transparency and efficiency to address these environmental concerns.
Penn State Offers Free Webinar on Farm AI Marketing
Penn State Extension is hosting a free webinar on using artificial intelligence for farm marketing. The event is scheduled for May 15 from noon to 1 p.m. It is designed for agritourism operators, farm market managers, and new farmers. Participants will learn how AI tools are used in everyday life and how to apply them to their own businesses. Attendees must register by noon on May 15 to get the link to the live session and the recording.
AI Tools Transform the Travel and Airline Industry
The airline and travel industry is rapidly integrating artificial intelligence into its operations. Booking.com CEO Glenn Fogel discussed how AI is changing the way people book and experience travel. These tools help streamline processes and improve customer interactions across the sector. The shift represents a major change for an industry that has traditionally relied on manual booking and scheduling systems.
Databricks Releases Tool to Fix Clinical Trial Data
Databricks launched a new open-source tool called the Site Feasibility Workbench to improve clinical trials. The platform unifies disconnected data sources that often cause delays in finding suitable research sites. It uses machine learning models trained on a sponsor's own historical data to predict site success. The system includes an audit trail for every prediction to meet strict medical regulations. This tool aims to reduce the time and money lost when trials fail to enroll enough patients.
UK Study Finds One in Seven Prefer AI Doctors
A new survey in the UK found that 14 percent of people would prefer to ask an AI chatbot for medical advice instead of seeing a doctor. The study was conducted by YouGov for the Royal College of GPs. The main reasons given were convenience, lower cost, and 24/7 availability. Doctors warn this trend is concerning and could create a two-tier healthcare system. Only 71 percent of respondents were aware of AI chatbots, but just 14 percent had actually used one for health questions.
OpenAI Competition Shows AI Agents Boost Research
OpenAI held a competition called Parameter Golf to test AI coding agents under strict limits. Participants had to minimize errors in a dataset while keeping their code under 16 megabytes and training time under 10 minutes. Over 1,000 people submitted more than 2,000 entries using AI agents to help them innovate. The event highlighted new techniques in model compression and optimization. OpenAI plans to run similar challenges in the future to discover new talent and research ideas.
Three Companies Lead the Future of Native AI
Three organizations are defining the next stage of artificial intelligence development. Thinking Machines focuses on low-latency human-AI collaboration in the Asia-Pacific region. The Effable movement is building interpretable safety frameworks to make AI decisions transparent. Isomorphic Labs is using AI to design new drugs by converting protein folding models into end-to-end engines. Together, they represent a shift from AI as a tool to AI as a core part of human and biological systems.
AI Benefits Expand Beyond Tech Giants and Investors
The value of artificial intelligence is spreading from tech companies to various industries and everyday life. Businesses use AI to detect fraud, analyze documents, and manage inventory more efficiently. Healthcare organizations apply it to medical imaging and drug research. Homeowners benefit from AI in energy management and smart appliances. Experts say the key to AI value is improving customer outcomes rather than just replacing human workers.
Sources
- Microsoft’s new AI system finds 16 Windows flaws, including four critical RCEs
- Microsoft’s agentic AI system found four critical Windows RCE flaws
- Microsoft's MDASH AI System Finds 16 Windows Flaws Fixed in Patch Tuesday
- Fortinet enhances FortiAIGate platform with NVIDIA accelerated AI security capabilities
- IBM Announces Red Hat AI Inference and Red Hat OpenShift Virtualization Service on IBM Cloud
- Datacentres using 6% of electricity supply in UK and US, research says
- Artificial Intelligence for Farm Marketing the Topic of Extension Webinar
- How AI is changing travel
- Databricks Unifies Clinical Data
- One in seven prefer consulting AI chatbots to seeing a doctor, UK study shows
- OpenAI's "Parameter Golf" Reveals AI's Role
- Architectural Interactivity, Linguistic Interpretability, and Molecular Synthesis: The Frontier of Native AI
- Who Benefits from Artificial Intelligence?
Comments
Please log in to post a comment.