Cybersecurity, AI, etc.
Artificial intelligence security (LLM/agentic system vulnerabilities, AI-enabled attacks, model security research, prompt injection, data poisoning)
GitLab has fixed CVE-2026-90970, described as a critical 9.9 flaw in AI Gateway that could allow logged-in Duo Agent Platform users to run commands on self-hosted gateways. Administrators should identify…
Threat researchers report that China-aligned TA419 targeted U.S. AI policy experts using reply-triggered adversary-in-the-middle phishing designed to capture Microsoft credentials and session cookies. Defenders should review identity-provider sign-in activity, enforce…
GitLab has warned customers to patch a critical vulnerability in its AI Gateway service after reporting that the flaw could allow arbitrary command execution on vulnerable instances. The issue puts…
OpenAI says it identified and disrupted a coordinated campaign designed to extract protected model reasoning at scale. The company attributed a core cluster of the activity to people associated with…
Threat actors are abusing ChatGPT Custom GPTs to make malicious destinations look like legitimate AI product offerings, according to reporting based on Huntress research. The campaign directs victims to a…
Anthropic has published a new assessment of Zhipu AI’s GLM-5.3, arguing that the open-weight model has crossed an important threshold in cyber capability. The report says GLM-5.3 can autonomously develop…
OpenAI has proposed safety cases for frontier AI training: structured evidence, incident-derived regression tests, postmortems, and disclosure after investigations conclude. Why it matters A safety case connects risk claims to…
OpenAI says GPT-6.1 Sol shows lower failure rates than GPT-6 Sol in evaluations of tool transparency, explicit restrictions, and avoidance of unauthorized agentic outcomes. Security implications Vendor evaluations are evidence,…
Zscaler’s ThreatLabz 2026 Ransomware Report says ransomware is rising and describes AI-assisted tactics, data theft, and targeting of privileged employees. How to use the finding The claims are vendor-report findings,…
A security advisory for the official Model Context Protocol Python SDK warns that affected OAuth client implementations can send OAuth credentials to an attacker-controlled token endpoint when they connect to…