Anthropic has alerted Claude users that attackers are using infostealer malware to steal login sessions, bypassing multi-factor authentication to drain paid credits. The company forced logouts, refunded unauthorized charges, and urged users to remove malware, emphasizing the issue stems from user-side infections, not Claude itself.
SlowMist said on July 29 that it has identified a newly discovered recruitment scam aimed at people working in Web3. In the campaign, attackers pose as recruiters and persuade targets to install an AI meeting application called "Relay," which is actually an information-stealing malware program. According to SlowMist, the malware is designed to target both macOS and Windows users and can steal browser credentials, crypto wallet-related data, Keychain data, and Telegram session information. The security firm said it has completed sample analysis and disclosed how the attack chain operates. SlowMist also warned users to be careful when joining online interviews or recruitment processes, avoid running unverified applications, and pay attention to unusual installation requests or system password prompts.
SlowMist said it has identified a newly discovered recruitment scam aimed at people working in Web3. In the campaign, attackers pose as recruiters and persuade targets to install an AI meeting application called "Relay," which is in fact an information-stealing malware sample. According to SlowMist, the malware targets both macOS and Windows users and is capable of stealing browser credentials, crypto wallet-related data, Keychain data, and Telegram sessions. The security firm said it has completed sample analysis and disclosed how the attack chain operates. SlowMist also urged users to stay cautious during online interviews and hiring processes, avoid running unverified applications, and watch for unusual installation requests or system password prompts.
The BBC has published an interactive test asking readers to tell real human faces from AI-generated ones, and its main takeaway is not that synthetic faces are flawless, but that people tend to overestimate their ability to spot them. That gap is turning deepfakes into a practical tool for social-engineering attacks tied to online hiring. The report says fake recruitment outreach has become an effective lure, especially for people working in tech, Web3, and AI. Attackers often pose as recruiters, well-known companies, or startup hiring teams on LinkedIn, X, and Telegram, offering attractive remote roles and salaries. Once a target agrees to interview, they may be asked to download an “interview app,” install a meeting tool, open a macro-enabled file, or run a coding test that actually contains malware or an infostealer. The BBC also cites research from the Australian National University showing that participants identified AI faces with an accuracy rate of about 41%. According to the report, today’s most convincing AI faces often look suspicious not because they appear broken, but because they seem too symmetrical, too polished, and too statistically perfect.
Europol said Operation Endgame’s latest phase froze more than €41 million in criminal crypto assets and dismantled infrastructure tied to SocGholish, Amadey, and StealC.
Security firm HiddenLayer reveals a fake OpenAI Privacy Filter model on Hugging Face hit #1 in 18 hours with 240K downloads, hiding a six-stage Rust infostealer targeting browser passwords, crypto wallet seed phrases, and SSH keys.
NJCCIC reveals a sophisticated attack using fake CAPTCHA pages that trick users into pasting malicious commands into Windows Run, silently deploying SectopRAT, Lumma, Vidar infostealers and rootkits via legitimate mshta.exe, targeting government employees and car dealership websites.
New Jersey cybersecurity agency warns of fake CAPTCHA pages that trick users into pasting malicious commands into Windows Run, deploying SectopRAT infostealer against government employees and auto dealership sites.
