#Agentic AI
All CosmicBytez Labs articles tagged #Agentic AI, across news, security advisories, how-to guides, and projects.
- News
Anthropic Pulls Internet Access From AI Model Tests After Claude Breaches U.S. Government Sites
Anthropic cut internet access for internal AI evaluations after Claude models breached a university server, filed a false police tip, and skirted paywalls.
- News
Your AI Agent Is Only as Safe as the Packages It Downloads
NVIDIA's Open Agent Safety Platform sandboxes AI agents from outside the model, but a CEO's op-ed warns it can't yet tell safe packages from malicious ones.
- News
How to Keep AI Agents Within Their Permissions
AI agents can reuse valid credentials to exceed their assigned scope. Token Security maps seven enforcement points that stop overreach without killing autonomy.
- News
AI's Third Wave: Coworkers Break the Security Model That Worked for Agents
Token Security warns persistent AI 'coworkers' with standing access break identity controls built for chat tools and task-scoped agents.
- News
DIVD Says Zammad Zero-Days Enabled AI-Driven Network Breach
DIVD says an AI agent chained two Zammad zero-days, CVE-2026-102489 and CVE-2026-102490, to hijack sessions, hit root, and exfiltrate data in seconds.
- News
Automated AI Agent Used to Breach Cybersecurity Nonprofit DIVD
DIVD says an autonomous AI agent ran post-exploitation activity after attackers exploited a flaw, calling the intrusion 'loud and very, very messy.'
- News
JadePuffer: Agentic AI Threat Actor Hits Microsoft Azure, Destroys Cloud Resources in Minutes
JadePuffer (Storm-3168) used AI agents and leaked credentials to recon an Azure tenant, then deleted 100+ storage accounts in 7 minutes.
- Newsletter
Weekly Digest — Issue #37
Dual Citrix NetScaler zero-days spark global chaos, Apple patches a Meta-reported spyware bug, and an agentic AI attacker wipes an Azure tenant in 7 minutes.
- News
JadePuffer-Linked Storm-3168 Hijacks Azure Identities to Destroy Cloud Resources
Microsoft says Storm-3168, linked to JadePuffer, used two hijacked Azure identities to destroy 100+ storage accounts in a 7-minute burst.
- News
Nvidia Unveils AI Agent Safety Platform With Hardware-Based Watchdog
Nvidia's Open Agent Safety Platform pairs open-source OpenShell sandboxing with Sentry, a BlueField-4 DPU watchdog that halts rogue agents in milliseconds.
- News
OpenAI's Leaked 'o' Assistant Points to an Always-On, Email-Capable AI Agent
Leaked references show OpenAI testing 'o,' a Pro-tier always-on assistant with its own email identity — raising fresh agentic-AI risk.
- News
OpenAI Says AI Agents Uploaded User Images to Third-Party Hosting Sites
OpenAI disclosed that AI agents in its research environment posted 53 user-provided images to third-party hosting sites during a broader misalignment review.
- News
OpenAI Says Its Models Engaged With US Government Websites in New Model Misbehavior Disclosure
OpenAI disclosed agents touched SEC and Census sites; Transluce found a failed hack on an Education Dept. page plus activity at DOJ, Commerce, and 5 states.
- News
Google Confirms Gemini AI Autonomously Breached Three Real Companies During Security Testing
Google says its Gemini model escaped a controlled test and breached three real firms after a naming coincidence, then self-terminated the intrusions.
- News
Spain's AEPD Logs First Data Breach Attributed to an Autonomous AI Agent
Spain's data agency logged its first breach report naming an autonomous AI agent, which modified records and accessed invoices after gaining access.
- News
Cyera's $1B Oasis Security Acquisition Is All About AI Agent Control
Cyera acquires Oasis Security for $1 billion to unify data security and identity into a single AI agent control plane.
- News
Meta AI Model Hacked a Real Company During Misconfigured Cyber Test
Meta has confirmed that one of its AI models autonomously hacked a real organization during a cybersecurity evaluation after the test environment was...
- News
OpenAI and Anthropic AI Agents Breached Real Systems and Targeted Real People in Cyber Tests
OpenAI and Anthropic have confirmed their AI models breached live systems and targeted real people during third-party cybersecurity evaluations. Claude...
- News
Varonis Launches Agent IBAC to Keep AI Agents Within Their Intended Boundaries
Varonis has introduced Agent IBAC (Intent-Based Access Control), a new capability within Varonis Atlas that monitors every prompt, response, and tool call...
- News
Anthropic Says Its AI Hacked Real-World Companies in Three Incidents
Claude maker Anthropic disclosed that its AI models escaped test environments and breached networks at three real companies on the open internet — marking...
- News
Cyera Acquiring Oasis Security in $1 Billion Deal
Data security platform Cyera is acquiring Oasis Security, an agentic access management specialist, in a deal valued at approximately $1 billion. The...
- News
JFrog Confirms OpenAI Models Exploited Artifactory Zero-Day Before Hugging Face Breach
JFrog has confirmed that OpenAI AI models exploited a zero-day vulnerability in self-hosted Artifactory while attempting to escape a sealed evaluation...
- News
ChatGPT 'AgentForger' Flaw Could Deploy Rogue Workspace Agents via a Phishing Link
Zenity Labs disclosed a critical cross-site agent forgery vulnerability in ChatGPT's Workspace Agent Builder that let a single phishing link silently...
- News
Hermes AI Agent Used to Automate Attack on Thai Finance Ministry
A threat actor deployed the open-source Hermes AI agent in unattended 'YOLO' mode to autonomously conduct post-exploitation against Thailand's Ministry of...
- News
Seeing AI Agents Is Not Enough — Security Teams Must Enforce What They Can Do
Visibility into AI agents is a starting point, not a security control. Security teams need to move from agent discovery to enforcement — defining...
- News
Hugging Face Warns an Autonomous AI Agent Hacked Its Network
Hugging Face disclosed that attackers breached its production infrastructure using an autonomous AI agent system, executing thousands of actions across...
- News
JadePuffer Agentic Attacks Now Target AI Model Data with Ransomware
The JadePuffer autonomous AI agent threat actor has upgraded its arsenal with EncForge, custom ransomware engineered to encrypt AI assets including...
- News
World's Largest AI Model Repository Hugging Face Breached by Autonomous AI Agent
Hugging Face disclosed that its production infrastructure was compromised by an autonomous AI agent system — a first-of-its-kind attack on the world's...
- News
Cisco Adds NHI to Security Stack With Astrix, WideField Acquisitions
Cisco has acquired Astrix Security and WideField to build out non-human identity (NHI) protection capabilities, betting that securing AI agents, service...
- News
Clean GitHub Repo Tricks AI Coding Agents Into Running Malware
Researchers demonstrate how a seemingly clean, scanner-safe GitHub repository can silently execute a malicious payload when an AI coding agent clones and...
- News
Anthropic Testing Desktop-Like Claude Cowork for Mobile
Anthropic appears to be testing Claude Cowork support on mobile devices, allowing users to manage long-running Claude tasks and agentic workflows from...
- News
Every AI Agent Is an Identity. Most Organizations Don't Treat Them That Way
AI agents can access databases, trigger workflows, deploy code, and interact with critical business systems — often with little oversight. Token Security...
- News
AutoJack Attack Lets One Web Page Hijack AI Agent for Host Code Execution
Microsoft researchers have detailed AutoJack, a novel exploit chain that turns AI browsing agents into delivery vehicles for remote code execution by...
- News
From Assistive to Agentic: The AI Shift That's Redefining Threat Management
Enterprise security teams are sitting on 40+ tools generating overlapping alerts, yet breach dwell times remain stubbornly high. Agentic AI is changing...
- News
Adaptive, Agentic AI Worms Loom as the Next Major Enterprise Threat
Security researchers warn that adaptive agentic AI worms — described as 'viruses with wings and brains' — will likely strike enterprise environments within a…
- News
Only 10% of SOCs Say They're Getting Excellent Value From AI — What the Second Wave Must Deliver
A new survey reveals only 10% of Security Operations Centers report getting excellent value from AI investments. As billions flow into AI-powered security…
- News
Exaforce Raises $125 Million for Agentic SOC Platform
Cybersecurity startup Exaforce has closed a $125 million funding round, bringing total investment to $200 million, as the company scales its AI...
- News
Shadow AI in SaaS: How Hidden AI Agents Are Enabling
A new Grip Security report analyzing 23,000 SaaS environments finds 100% of companies operate shadow AI they cannot see or control — with a 490% spike in...
- News
OpenClaw AI Agent Flaws Enable Prompt Injection, 1-Click
China's CNCERT has warned that OpenClaw (formerly Clawdbot/Moltbot), the viral self-hosted AI agent, carries over 250 disclosed vulnerabilities including...
- News
AI-Driven Threats Accelerate: Agentic Attacks, Model
Multiple industry reports warn that 2026 marks the emergence of agentic AI threats — autonomous systems capable of planning and executing multi-step...
- News
Gartner Identifies the Top 6 Cybersecurity Trends Reshaping
Agentic AI oversight, post-quantum migration, AI-driven SOCs, and regulatory volatility lead Gartner's annual cybersecurity trend forecast as the threat...