Investigators at Cedars-Sinai's Smidt Heart Institute in Los Angeles developed an artificial intelligence (AI) method that effectively detects atrial fibrillation, a condition characterized by abnormal heart rhythms, from echocardiogram imaging. The deep learning model was trained with over 100,000 echocardiogram videos to distinguish between normal heart rhythms and those indicating atrial fibrillation. Remarkably, this AI can identify patients likely to develop or who have had atrial fibrillation within 90 days, surpassing traditional risk estimation methods. This advancement highlights the potential for AI to play a significant role in early cardiac care, especially for conditions like atrial fibrillation that are challenging to diagnose due to their intermittent nature. The technology promises to improve early detection and treatment, potentially preventing serious cardiovascular events in patients with this often undetected arrhythmia.

Cedars-Sinai research shows deep learning model could improve AFib detection Healthcare IT News

Zscaler addressed rumors of a system breach by confirming that an isolated test environment, unconnected to its main infrastructure or customer data, was exposed online and subsequently taken offline for forensic analysis. The company's ongoing investigation has found no evidence of any compromise to its core customer or production environments. These rumors began circulating after a threat actor named IntelBroker claimed to be selling access to a cybersecurity company's sensitive data. Despite these claims, Zscaler reassures that its primary operational and customer data environments remain secure, indicating no impact from this incident on company or customer data.

Zscaler takes "test environment" offline after rumors of a breach BleepingComputer

In a conversation with MIT Technology Review, OpenAI CEO Sam Altman outlined a future where AI evolves beyond its current capabilities to act as highly competent, personal agents involved in various aspects of our lives without the need for new hardware or vast amounts of additional training data. Altman described these AI systems as super-competent colleagues that can manage both simple and complex tasks independently, transforming our interaction with technology. He emphasized that despite the potential for new AI-specialized devices, the envisioned applications could operate effectively in the cloud, suggesting that existing devices might suffice. Altman also touched on the future of artificial general intelligence (AGI), predicting the development of multiple versions of AGI, each with different strengths.

Sam Altman says helpful agents are poised to become AI’s killer function | MIT Technology Review MIT Technology Review

This article details a novel network technique discovered by researchers that allows an attacker to bypass VPN encapsulation via DHCP (Dynamic Host Configuration Protocol), effectively forcing a target user’s traffic outside of their VPN tunnel. Termed as "decloaking," this method subtly exploits the DHCP without disrupting the VPN's control channel, leaving the user unaware as their data transmissions are not encrypted by the VPN. Despite attempts to inform affected parties, the technique—which is believed to be exploitable since 2002—remains a significant threat. The article emphasizes the difficulty in mitigating this vulnerability due to the essential role DHCP plays in network connectivity and suggests the implementation of network namespaces as a potential fix for systems that support it, like Linux. The research aims to raise awareness within the security community about this threat and the challenge in notifying every VPN provider and user, hoping for wider implementation of effective countermeasures.

TunnelVision (CVE-2024-3661): How Attackers Can Decloak Routing-Based VPNs For a Total VPN Leak Leviathan Security