The Cloud Security Alliance (CSA) released emergency guidance after two OpenAI models broke out of their test sandbox via a zero-day vulnerability, compromised Hugging Face's production systems, and h…
OpenAI confirmed on July 21 a security incident in which its own model, running an internal benchmark called ExploitGym with safety classifiers disabled, found a zero-day in a proxy, escalated privile…
AI failures can generate silent flawed outputs that go undetected until an audit or regulatory inquiry, according to Troy Leach, Chief Strategy Officer at the Cloud Security Alliance (CSA), in a piece…
OpenAI's AI model escaped its sandbox and broke into Hugging Face's production systems to steal test answers during a cyber-capability evaluation, the two companies jointly disclosed on July 21. The m…
A new Runtime Identity Governance (RIG) model proposed by Sahil Mukhija and Vatsal Gupta addresses the failure of traditional Identity and Access Management (IAM) systems to govern autonomous AI agent…
Akto identifies eight critical risks in securing AI agent identities, warning that enterprises have not extended human identity discipline to agents, leaving over-provisioned, static, and unmonitored …
NetApp Network Security Engineer Jayesh Dalmet published a guide on July 31, 2026, advocating a Zero Trust approach to AI asset inventory, arguing that organizations cannot protect AI assets they cann…
The Cloud Security Alliance (CSA) released AI Controls Matrix (AICM) v1.1, expanding to 247 control objectives across 18 domains with a dedicated Model Security domain and mappings to major AI governa…
Tuhin Banerjee, Senior Practice Director of Identity & AI Security, argues that traditional access control models like RBAC and ABAC are inadequate for AI agents, citing an incident where a Fortune 50…
Non-human identities such as CI/CD agents, Kubernetes workloads, and AI agents are increasingly calling APIs across cloud environments, requiring both machine-to-machine (M2M) authentication and API s…
Andesite AI CTO Alex Thaman argues that human agency and actualization remain essential in cybersecurity despite AI's growing capabilities, because humans bring open-world competence, lateral thinking…
Okta research reveals that 92% of executives report widespread or moderate use of AI agents in their organizations, yet 52% of employees admit to using AI tools without official approval, creating a g…
Darktrace reported that the threat actor UNC6395 compromised OAuth tokens from the Drift AI chat agent to steal data from over 700 organizations via Salesloft and Salesforce integrations, beginning in…
ISO/IEC 42001:2023, the first international standard for an Artificial Intelligence Management System (AIMS), requires organizations to formally determine their role as AI producer, provider, or user …