Full Summary
This Wednesday morning, Microsoft's new AI security system, MDASH, has identified 16 previously unknown vulnerabilities in Windows, including four critical remote code execution flaws. Both PCMag and Thurrott.com confirm MDASH uses over 100 specialized AI agents and multiple AI models to discover and prove exploitable bugs, outperforming other AI models with an 88.45% score on the CyberGym benchmark. The development comes as the cybersecurity landscape grapples with the rapid advancement of AI. Many organizations, according to Help Net Security and TechRadar, are struggling to secure AI systems, especially non-human identities like AI agents and automated workflows, which often operate with broad, unsupervised access. This creates a significant "AI identity risk gap." Dell and Intel, at their "Securing the AI Factory" event, highlight that 85% to 90% of AI projects are halted mid-implementation because security isn't integrated from the start. Meanwhile, companies are rushing to provide solutions. Exaforce secured $125 million to expand its AI-native security platform, and Saviynt appointed new leadership to boost its AI identity security. NetSPI launched AI-powered continuous pentesting, while Sweet Security introduced "Sweet Attack," an AI agent that simulates adversaries to fortify cloud defenses. Even consumer-focused security is evolving, with Gen Digital updating its Norton Neo AI browser with new privacy and security features. But then, what nobody expected: the Vercel breach, detailed by Cybersecurity Insiders, shows the danger of forgotten AI app trials. A single OAuth token from an employee's old trial of a deprecated AI product, Context.ai, led to a supply chain attack, granting attackers access to internal dashboards and sensitive data. This highlights how easily "shadow AI" can create invisible bridges for attackers. This means your personal data and the security of the applications you use daily are at greater risk from both sophisticated AI-driven attacks and overlooked vulnerabilities in how AI tools are adopted.