Startup Strips Refusal Behavior From AI Models and Profits From It
Abliteration AI sells API access to an uncensored version of Z.AI's open-weight model GLM-5.3, without standard identity verification, shifting responsibility …
All articles in the "Cybersecurity & Protection" category.
Abliteration AI sells API access to an uncensored version of Z.AI's open-weight model GLM-5.3, without standard identity verification, shifting responsibility …
The ownCloud flaw CVE-2023-49105, known since 2023, is now being actively exploited; affected versions should be urgently updated to 10.13.1 or later.
Germany's federal government drops the constitutional amendment promised in the coalition agreement to strengthen the BSI, relying instead on existing cooperat…
A buffer overflow in the Linux kernel's IPv6 stack (CVE-2026-53362) is already being actively exploited, letting unprivileged users corrupt memory via UDPv6 so…
The critical Entra ID flaw CVE-2026-69836, rated CVSS 10.0, was not actively exploited despite circulating media reports; Microsoft had corrected its initial a…
Several vulnerabilities in Firefox and Thunderbird allow privilege escalation, data manipulation and in some cases malicious code execution.
Several new vulnerabilities in Google Chrome allow malicious code execution, privilege escalation and access to confidential data.
Germany's new AI Safety and Security Institute, led jointly by the BSI and the Federal Network Agency, is meant to assess the risks of capable AI models, thoug…
Anthropic proactively logs out Claude users and removes stored payment methods after infostealers like Vidar and Lumma abused stolen login sessions.