A new Android backdoor called RatHat utilizes an AI-powered system to remotely navigate compromised devices, while also stealing sensitive information and using a variety of methods to maintain its presence. Researchers at Zimperium’s zLabs found indications that RatHat may be associated with threat actors based in China after they discovered Chinese language prompts within the malware’s AI subsystem.  The malware is reported to be distributed through malicious advertising, SMS messages and phishing websites that promote APK downloads outside Google Play. RatHat takes advantage of Android’s Accessibility permissions to obtain extensive control over infected devices. This allows it to enable Developer Options and Wireless Debugging, granting it a local shell-level execution environment without the need for a separate computer. Researchers observed similarities with this technique that have been previously seen in the ToxicPanda and RedHook Android malware families.  The malware utilizes Android Debug Bridge (ADB) access to install a Go-based agent called liblocal-service.so. The agent can execute commands with ADB shell privileges, bypassing battery restrictions and establishing persistence. It can also restore the malware in the case that the main component is removed or stopped. The relationship works in both directions, with the malware being able to restore the agent if the agent itself is deleted.  RatHat also makes use of a second component, libmedia_codec.so, which acts as an FRP reverse-proxy client and establishes a persistent tunnel to the attackers. The malware has the ability to display HTML overlays over targeted banking and cryptocurrency applications in order to acquire the users’ credentials. Its information-stealing capabilities include SMS messages and notifications, including one-time passwords. RatHat can also monitor text changes, extract URLs from browser address bars and capture lock-screen PINs, passwords and unlock patterns.  One of RatHat’s most interesting features is its AI-powered interface automation engine. According to Zimperium, the malware converts the Android Accessibility tree into XML and sends the resulting information to an unnamed popular AI assistant. This system can recognize the screen coordinates of requested interface elements and determine their displayed text and provide navigation commands such as scrolling instructions. This enables the malware to navigate Android interfaces more dynamically than other malware that exclusively rely on predetermined scripts.  Zimperium stated that the AI-driven system makes the malware more adaptable and arguably harder for security software to detect. RatHat actively prevents victims from uninstalling the malware. When an uninstall confirmation screen appears, the malware can intercept the process and cancel the removal and display a fake Google Play overlay, which shows a fraudulent error message. The malware also contains a number of anti-analysis measures, including APK container manipulation, an unusually large 61MB Android manifest and invalid DEX pseudo-instructions that are designed to confuse or disrupt the functionality of security analysis tools.  Android users are advised to avoid downloading APK files from outside Google Play unless the publisher is explicitly trusted and to be careful when granting Accessibility permissions to applications. In addition, users should regularly scan their devices using Google Play Protect.

Google researchers observed threat actors using AI-developed zero-days, advanced reconnaissance, and autonomous malware.

Experts At The Table: From jail breaking an AI to security and integrity of AI training data, what are the best ways to fend off threats from AI-based attacks.

Michael Engle, Co-Founder and CSO at 1Kosmos, explains steps organizations can take to strengthen identity proofing.

Researchers warn that a newly identified open-source AI security testing platform called CyberStrikeAI was used by the same threat actor behind a recent campaign that breached hundreds of Fortinet…

 CrowdStrike Holdings Inc. released a new report earlier this month that illustrates how cyber threats evolved significantly in 2024, with attackers pivoting towards malware-free incursions, AI-assisted social engineering, and cloud-focused…

Google DeepMind's Apr 2, 2026 paper identifies six AI-agent attack categories — including invisible HTML commands and multi-agent flash crashes — that could reshape vendor spend.

A memory injection attack dubbed Minja turns AI chatbots into unwitting agents of misinformation, requiring no hacking and just a little clever prompting. The

According to Anthropic's September 2025 disclosure, a state-sponsored campaign successfully automated 80-90% of an intrusion using their agentic AI tool, Claude Code. This isn't a story about AI helpi...

Thursday. Another week, another batch of things that probably should've been caught sooner but weren't. This one's got some range — old vulnerabilities getting new life, a few "why was that even possible" moments, attackers leaning on platforms and tools you'd normally trust without thinking twice. Quiet escalations more than loud zero-days, but the kind that matter more in

Google says China-linked hackers are running AI inside compromised cloud environments, using victims’ computing power while reducing outside monitoring.

AI is creating far more detection noise for CrowdStrike than human-triggered incidents, with millions of daily leads and tens of thousands of customer alerts over the past year. The report says AI is now both a weapon and a target, accelerating vulnerability exploitation, supply chain attacks, and threats against AI systems themselves. #CrowdStrike #TeamPCP

AI has changed how fast attacks move. Work that once took an attacker days now takes minutes. Using models like Mythos, attackers write tailored bait, pick targets, test what lands, and jump to the next host before your team clears the first alert. That is the gap, and it is not your fault. The tools and runbooks most teams run on were built for attackers who work at human speed. AI-driven

AI is changing cyber risk by speeding up reconnaissance, social engineering, malware adaptation, and decision-making, which compresses the time defenders have to respond. Organizations must update governance, identity verification, detection, and incident response to handle AI-speed attacks. #Sygnia #GuySegal

The article discusses the growing threat of AI-powered cyberattacks, emphasizing their ability to bypass traditional security measures and exploit human vulnerabilities. It highlights the need for robust cybersecurity strategies, including advanced threat detection, employee training, and multi-layered defenses. The piece also explores the ethical implications of AI in warfare and surveillance, warning of potential misuse by state and non-state actors. Experts stress the importance of international cooperation to address these challenges, as AI-driven attacks can target critical infrastructure, financial systems, and personal data. The article concludes with recommendations for individuals and organizations to adopt proactive measures, such as regular system updates, encryption, and incident response planning, to mitigate the risks posed by evolving cyber threats.

Bots now account for more online shop traffic than real humans

Monday hit like a cron job with anger issues. A busted auth path here, a repo-side faceplant there, some "patched-ish" thing already getting chewed on in the wild, and then the usual bonus round: poisoned dev tools, sketchy forum chatter, phishing kits pretending to be productivity, and AI lowering the bar for people who already thought 'curl | sh' had a personality. The vibe is simple: old

Thomas Le Coz, CEO and Co-Founder, Arsen Security, explains why social engineering attacks are outpacing training and exposing gaps in human response.