Daily Security Intelligence Briefing — July 28, 2026

1. ChatGPT AgentForger Flaw Could Deploy Rogue Workspace Agents via Phishing - Source: https://cybersecuritynews.com/chatgpt-agentforger-vulnerability/?utm_source=tldrinfosec - Technical Severity / Domain: High / AI Agent & Workspace Security - Why You Must Read This: Zenity Labs discovered that ChatGPT Workspace Agent Builder accepted initialization states via URL parameters (template_name, initial_assistant_prompt) and auto-submitted prompts upon page load. A single malicious link sent to a logged-in user with pre-authorized connectors could silently create a rogue agent, bind existing Outlook, Gmail, Slack, Drive, and Teams integrations, and flip write action permissions from 'Always ask' to 'Never ask' without prompting an OAuth consent screen. The deployed agent ran every five minutes to check an attacker inbox, exfiltrate files, and execute tasks. - Recommended Immediate Action: Audit all active ChatGPT Workspace Agents, revert write-action approval settings to 'Always ask', and inspect deep links in emails attempting automated agent creation.

2. Hackers Hijack Hotel Wi-Fi DNS to Steal Microsoft 365 Accounts - Source: https://links.tldrnewsletter.com/9NXQwz - Technical Severity / Domain: High / Cloud Security & Network Infrastructure - Why You Must Read This: ReliaQuest identified an active phishing campaign where threat actors modify hotel Wi-Fi DNS settings to redirect Microsoft 365 authentication requests to attacker-controlled phishing portals. The attack leverages OAuth device code authentication flows, tricking users into approving device login requests and bypassing traditional MFA controls. - Recommended Immediate Action: Disable device code authentication flows in Microsoft Entra ID via Conditional Access policies and transition remote employees to FIDO2 phishing-resistant authentication.

1. Microsoft Unveils Project Perception Agentic Defense System - Subject / Topic Summary: Microsoft introduced Project Perception, a multi-model agentic security platform entering public preview on August 3, 2026. - Key Technical & Strategic Takeaways: Coordinates red-team agents (attack path discovery), blue-team agents (risk investigation), and green-team agents (automated remediation) across identity, endpoint, cloud, and AI environments to deliver continuous autonomous defense while keeping security teams in control. - Essential Direct Quote: "Microsoft's Project Perception is an agentic security system that coordinates red-team agents to identify attack paths, blue-team agents to investigate risks, and green-team agents to take corrective action across identities, endpoints, applications, data, clouds, and AI systems." (Source: https://blogs.microsoft.com/blog/2026/07/27/rethinking-security-for-the-age-of-ai/?utm_source=tldrit)

2. NIST Releases Draft SP 800-239 on AI Data Center Security Analysis - Subject / Topic Summary: NIST published Draft Special Publication (SP) 800-239, "AI Data Center Security Analysis: A High-Performance Computing (HPC) Driven Approach". - Key Technical & Strategic Takeaways: Evaluates threat models and security gap analyses specifically tailored for AI model training and inference hardware/software stacks, contrasting AI data center architectures with traditional HPC systems. - Essential Direct Quote: "NIST invites public comments on the initial public draft (ipd) of Special Publication (SP) 800-239... which delivers a thorough threat and security gap analysis for purpose-built AI infrastructure used in model training, inference, and applications." (Source: https://csrc.nist.gov/pubs/sp/800/239/ipd)

3. BlueNoroff Fake Meeting Kit Targets Webcams, Browser Wallets, and Mac Credentials - Subject / Topic Summary: Technical breakdown of North Korean threat group BlueNoroff (TAG-195) using deepfake meeting portals to compromise enterprise devices. - Key Technical & Strategic Takeaways: Cloned Zoom/Teams portals open webcams and enumerate Web3 wallets via EIP-6963. Fake SDK updates execute clipboard PowerShell loaders (Trojan.NukeSped), block Windows Defender, and harvest Telegram session data from IndexedDB. - Essential Direct Quote: "Defensive actions include alerting on Defender exclusions, monitoring clipboard PowerShell, hunting processes accessing IndexedDB, flagging unknown contacts before SDK updates, and blocking 11 C2 domains." (Source: https://gbhackers.com/bluenoroff-fake-meeting-kit/?utm_source=tldrinfosec)

4. Unauthenticated API Endpoints Exposed Fleet Telematics Platform - Subject / Topic Summary: Vulnerability analysis of VE Commercial Vehicles' My Eicher platform exposing 2M customer records and vehicle controls. - Key Technical & Strategic Takeaways: Open API endpoints exposed full OTP histories since 2021, enabling account takeover by matching mobile numbers to OTPs, full fleet tracking, and exposure of 76K identity documents before remediation. - Essential Direct Quote: "By walking the API path, Eaton found open endpoints that returned user records, encrypted passwords, and a full OTP history since 2021, enabling account takeover by matching mobile numbers to OTPs or resetting passwords via API." (Source: https://eaton-works.com/2026/07/27/my-eicher-hack/?utm_source=tldrinfosec)

Tier 3 — Industry Noise & Awareness Only

• - Tribeca Film Festival Data Exposure: Unencrypted database exposed PII and bcrypt password hashes for 163K users and celebrities (https://www.expressvpn.com/blog/tribeca-film-festival-data-exposed/?utm_source=tldrinfosec).

• - GitHub Issues $100K Bounty for RCE: GitHub paid a $100K bounty for CVE-2026-3854, an unauthenticated remote code execution flaw in repository URL handling (https://runtimewire.com/article/github-issues-100-000-bounty-for-critical-rce-vulnerability-disclosed-by-sagitz?utm_source=tldrinfosec).

• - Dependabot & PyPI 72-Hour Hold: Dependabot introduced a default 72-hour delay on version update pull requests to prevent supply chain compromises from newly published malicious packages (https://links.tldrnewsletter.com/IN8OtM).

• - Global IT Spending Reaches $6.37T: Enterprise tech spending reaches $6.37T as vendors bundle AI costs into subscriptions, drawing CIO pushback (https://www.theregister.com/ai-and-ml/2026/07/27/tech-sector-pours-1t-into-ai-and-sends-customers-the-bill/5278845?utm_source=tldrit).

• - grokpatrol Utility Released: Open-source tool detects Grok Build CLI installations and flags git secret uploads to xAI (https://github.com/optimuslabs-io/grokpatrol?utm_source=tldrinfosec).

• - zeedumper Kubernetes Tool: Tool connects to Kubernetes clusters to dump component z-pages via API server proxy (https://github.com/raesene/zeedumper?utm_source=tldrinfosec).