AI reads 36 sources every 2 hours and ranks each story by importance from 1 to 5 - so you see what actually matters.
Anthropic has reported that its Claude AI model was misused in foreign missile development projects and international espionage operations. The disclosure highlights how advanced commercial AI systems can be repurposed for military and intelligence work despite usage restrictions. It underscores growing concerns over safeguards as frontier AI capabilities spread.
The Hacker News roundup covers 200 Android flaws, browser-based phishing, and 119,000 scam shops, plus 23 more security stories. It matters because it offers a broad snapshot of active threats facing mobile users, web users, and online fraud victims.
Researchers found that attackers exploited Google Play's Early Access program to distribute thousands of deceptive Android apps before they could be properly reviewed. The apps used the pre-release slot to evade detection and reach users at scale. The finding highlights a gap in Google's app vetting process and the risks of pre-release distribution channels.
Check Point has disclosed two critical vulnerabilities in its VPN products, both rated 9.8 out of 10, that allow attackers to execute remote code without authenticating. Flaws of this severity in perimeter VPN appliances are especially dangerous because such devices are typically internet-facing, giving attackers a potential entry point into corporate networks. Organizations running affected Check Point VPN gateways should apply the vendor's fixes or mitigations promptly.
Attackers deployed hundreds of autonomous AI agents to compromise more than 440 exposed PaperCut print-management servers, according to The Hacker News. The campaign marks a notable escalation in the use of AI-driven automation to scan, exploit, and move through victim networks at scale. Organizations running internet-facing PaperCut instances are urged to patch and audit for signs of intrusion.
Security researchers at The Hacker News report that a new Android banking trojan called Gigabud abuses Android's Work Profile feature to conceal itself from the security checks banking apps run to detect malware. By operating inside a separate managed profile, the malware can avoid scrutiny that would normally flag it on the device. The technique matters because it undermines a common defense used by banking apps to block fraudulent access to accounts.
Security commentator Bruce Schneier argues that AI tools are compressing the time attackers need to turn a disclosed software vulnerability into a working exploit. That squeezes the patch window for defenders, who must now fix flaws faster than ever. The shift matters because it erodes a longstanding assumption that organizations have weeks of breathing room after a flaw goes public.
The U.S. Cybersecurity and Infrastructure Security Agency added actively exploited flaws in Cisco, Citrix, and Fortinet products to its Known Exploited Vulnerabilities catalog. CISA set a Sept. 12 deadline for federal agencies to patch them, citing risks to widely used network and remote-access systems. The directive highlights urgent exposure for organizations using affected products.
A scan of internet-exposed LiteLLM proxy gateways found that roughly 10% still accepted the example admin key "sk-1234" shipped in the project's documentation. LiteLLM is widely used to route and manage access to multiple LLM providers, so a valid admin key can expose API credentials, model access, and potentially allow abuse of paid services. The finding underscores how often default example secrets survive into production deployments.