Definition

Cybersecurity encompasses practices, technologies, and processes designed to protect systems, networks, and data from unauthorized access, attacks, and damage.

Key Concepts

  • Threat Actors: Malicious entities (hackers, cybercriminal groups, state-sponsored actors)
  • Attack Vectors: Methods used to exploit vulnerabilities (phishing, supply chain attacks, exploits)
  • OAuth Security: Authentication protocol vulnerabilities when integrating third-party services
  • Supply Chain Attacks: Compromising trusted dependencies or integrations to gain access to target systems
  • Zero-Day Vulnerabilities: Unknown flaws exploited before vendors can patch (see zero-day-vulnerabilities)
  • Authorized Intent Chain: Attacks where every step is authorized, bypassing EDR/firewalls (authorized-intent-chain)

2026 Notable Incidents

Outsider Enterprise Phishing (June 2026)

  • google sued China-based outsider-enterprise for alleged gemini-assisted phishing
  • Scale: 2.5M scam texts, 9K fake sites in two weeks (May 2026); PhaaS at $88/week
  • Coordination: FBI + AT&T/T-Mobile/Verizon text blocking
  • Lesson: Frontier AI lowers cost of phishing site generation; fragmented abuse detection across providers

Agentjacking (June 2026)

  • Attack type: agentjacking via fake sentry error injection → model-context-protocolai-coding-tools
  • Impact: 2,388 exposed orgs; 85% success rate in controlled tests (Claude Code, cursor, codex)
  • Bypass: EDR, firewalls, IAM, VPNs — no unauthorized actions detected
  • Lesson: MCP tool output must be treated as untrusted input; runtime execution controls required

npm v12 Security Defaults (June 2026)

  • npm v12 blocks install scripts, Git deps, remote URL deps by default (July 2026)
  • Response to miasma-worm, Shai-Hulud, and 455K+ malicious packages in 2025
  • npm approve-scripts allowlist model shifts ecosystem to explicit permission

Miasma Worm Campaign (June 2026)

  • miasma-worm compromised Red Hat npm (30+ packages) and Microsoft GitHub (73 repos)
  • Valid SLSA attestations on malicious packages; AI agent auto-execution triggers

Vercel Breach (April 2026)

  • Attack type: OAuth-based supply chain attack via Context.ai
  • Threat actor: ShinyHunters
  • Method: Lumma Stealer malware exfiltrated OAuth tokens
  • Impact: Non-sensitive environment variables compromised
  • Lesson: Third-party OAuth permissions create “Identity Supply Chain” risk

Claude Mythos Discovery (May 2026)

  • anthropic’s Claude Mythos discovered thousands of zero-day vulnerabilities
  • 27-year-old OpenBSD flaw, 16-17 year-old FFmpeg vulnerability
  • Demonstrates AI can find previously undetectable security issues

July 2026 Cryptanalysis Milestone

July 2026 Agentic Security Stack

Sources