Google has issued a stark warning that threat actors are now using artificial intelligence to autonomously create working zero-day exploits, marking a significant escalation in offensive cyber capabilities. The AI models were observed successfully generating exploit code for previously unknown vulnerabilities, dramatically reducing the time from discovery to weaponization. In parallel, North Korean hacking groups have developed a novel technique weaponizing Git hooks to deploy cross-platform malware that infects developer environments across Windows, Linux, and macOS. By injecting malicious scripts into the pre-commit and post-merge hooks of Git repositories, the attackers ensure their payload runs every time code is committed or synced, enabling persistent access to source code, credentials, and CI/CD pipelines.
ODINI Malware Breaches Faraday-Shielded Systems via CPU Magnetic Emissions
Researchers have unveiled ODINI, a strain of malware that can exfiltrate data from air-gapped computers protected by Faraday cages. ODINI modulates the CPU’s magnetic field emissions generated during normal operations to encode stolen data, which is then received by a nearby smartphone or IoT device acting as a magnetic receiver. This side-channel technique bypasses physical isolation, network firewalls, and even electromagnetic shielding by leveraging the inherent magnetic leakage from processor cores during controlled workloads. The attack does not require any network connection, USB port access, or direct physical contact with the target machine.
Critical Ollama Memory Leak and Microsoft Patch Tuesday Remediate Massive Attack Surface
A critical memory leak vulnerability in Ollama, an open-source large language model deployment framework, has been disclosed, potentially exposing over 300,000 Ollama servers worldwide to remote data theft. The flaw leaks inference session data and sensitive model prompts from adjacent tenants on shared infrastructure. Meanwhile, Microsoft’s May 2026 Patch Tuesday addressed 120 vulnerabilities, including 29 critical remote code execution (RCE) flaws affecting Windows, Office, and the .NET framework. Administrators are urged to prioritize patching Ollama instances and the Microsoft RCE bugs (including CVE-2026-29145) as attackers are already developing exploits combining the AI-generated zero-day techniques with ODINI-style exfiltration methods.
Sources: https://cybersecuritynews.com/ai-zero-day-exploit/ , https://cybersecuritynews.com/bitunlocker-downgrade-attack-on-windows-11/ , https://cybersecuritynews.com/claudes-chrome-extension-vulnerability/ , https://cybersecuritynews.com/darkmoon-penetration-testing-platform/ , https://cybersecuritynews.com/fortinet-enterprise-products-vulnerabilities/ , https://cybersecuritynews.com/fortinet-fortisandbox-vulnerability/ , https://cybersecuritynews.com/ghostlock-attack/ , https://cybersecuritynews.com/hackers-hijack-microsoft-teams/ , https://cybersecuritynews.com/hackers-use-weaponized-jpeg-file/ , https://cybersecuritynews.com/ivanti-patches-multiple-vulnerabilities/ , https://cybersecuritynews.com/lets-encrypt-halts-certificate-issuance/ , https://cybersecuritynews.com/microsoft-365-copilot-vulnerabilities-data/ , https://cybersecuritynews.com/microsoft-patch-tuesday-may-2026/ , https://cybersecuritynews.com/new-stealthy-vidar-stealer-campaign/ , https://cybersecuritynews.com/no-blind-spots-how-top-mssps-prevent-incidents-withlive-threat-visibility/ , https://cybersecuritynews.com/north-korean-hackers-weaponize-git-hooks/ , https://cybersecuritynews.com/odini-malware-air-gapped-computers/ , https://cybersecuritynews.com/ollama-vulnerability-exposes-servers/
Source: Cyber Security News
