Latest / Neural Newscast / Storm-0940 Breaches 2,250 Cloud Tenants with AI Agents [Prime Cyber Insights]
Transcript
- [00:00] Announcer: From Neural Newscast. This is Prime Cyber Insights, [00:03] Announcer: intelligence for defenders, leaders, and decision makers. [00:11] Announcer: Welcome to Prime Cyber Insights. I am joined by [00:14] Announcer: Lauren to examine the narrowing window between vulnerability disclosure and active exploitation. [00:20] Aaron Cole: We begin today with Microsoft's confirmation that the China-linked group Storm 0940 [00:27] Aaron Cole: has breached over 2,250 cloud tenants by utilizing an open-source AI tool called Art. [00:35] Aaron Cole: The group automated the attack chain, achieving [00:37] Aaron Cole: privileged cloud access in under three hours in several instances. Aaron, [00:42] Announcer: that speed aligns with a broader industry trend. Rapid7's Q3 report notes that high severity flaws [00:49] Announcer: are now exploited within an average of three days following disclosure. [00:53] Announcer: We are seeing this currently with CVE-20260257, an authentication bypass in [00:58] Announcer: Palo Alto Networks Global Protect that. [01:01] Announcer: That has been under active exploitation by ransomware groups since June. [01:05] Aaron Cole: It is a race against time. Rapid7 highlights that while AI is automating the tactical middle [01:10] Aaron Cole: of attacks, specifically target triage, only eight percent of incidents currently use AI [01:16] Aaron Cole: as a core operational component. Human actors [01:19] Aaron Cole: are still selecting the targets, Erin. But AI is making their execution significantly faster. [01:25] Announcer: Turning to global enforcement, Interpol has concluded Operation Serengeti III. [01:30] Announcer: This massive sweep across 21 countries resulted in 1,462 arrests and the recovery of $105 million. [01:39] Announcer: They dismantled nearly 11,000 malicious networks, [01:42] Announcer: targeting threats ranging from business email compromise to illegal crypto mining in Angola. [01:47] Aaron Cole: That scale of disruption is essential as vulnerability volume grows. Erin, [01:53] Aaron Cole: OpenAI is now proposing a five-stage disclosure framework for critical software. [01:59] Aaron Cole: This follows their use of GPT 5.6 Sol to identify vulnerabilities in PHP. [02:06] Aaron Cole: The objective is to engage governments [02:08] Aaron Cole: and vendors early to build defenses before AI-assisted attackers can react. [02:13] Announcer: Regarding specific vulnerabilities, Signal users should review 2026 Cape Verde Scudo's 113,380. [02:22] Announcer: A flaw in the RingRTC library allows a remote peer to flood the event queue, [02:27] Announcer: preventing a call from properly disconnecting. In practice, Lauren, it means the other party [02:32] Announcer: could potentially continue to monitor your audio even after you believe you have hung up. [02:38] Aaron Cole: That is a significant privacy risk. The fix is included in Ring RTC version 2.72.0, [02:45] Aaron Cole: so immediate updates are mandatory. Whether it is nation-state espionage or a library flaw, [02:52] Aaron Cole: the common thread [02:53] Aaron Cole: is the requirement for rapid automated defense to match the current speed of the adversary, Aaron. [02:59] Announcer: Precision in patching is no longer optional; [03:02] Announcer: it is the baseline for security. That concludes our technical briefing for today. [03:07] Aaron Cole: Thank you for joining us in the briefing room. For more practitioner-focused analysis, [03:11] Aaron Cole: visit pci.neuralnewscast.com. Neural Newscast is AI-assisted, [03:17] Aaron Cole: human-reviewed. View our AI transparency policy at neuralnewscast.com. [03:23] Lauren Mitchell: This has been Prime Cyber Insights on Neural Newscast: [03:26] Lauren Mitchell: Intelligence for Defenders, Leaders, and Decision Makers.