{
  "id": 918205,
  "title": "GLM-5.3 is here with advanced cyber capabilities — and reportedly already found a 'serious vulnerability' in Cursor",
  "url": "https://urgent.news/2026/08/14/glm-5-3-is-here-with-advanced-cyber-capabilities-and-reportedly",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-14T22:56:16.000Z",
  "source": {
    "name": "VentureBeat",
    "slug": "venturebeat",
    "url": "https://venturebeat.com/technology/glm-5-3-is-here-with-advanced-cyber-capabilities-and-reportedly-already-found-a-serious-vulnerability-in-cursor"
  },
  "original_language": "en",
  "account": "Chinese AI startup Z.ai recently unveiled its latest large language model, GLM-5.3, which boasts significant improvements in coding and cybersecurity capabilities. GLM-5.3, available through the company's GLM Coding Plan and ZCode coding environment, represents a substantial upgrade over its predecessor, GLM-5.2, with enhancements derived solely from post-training scaling on diverse tasks and reinforcement-learning compute. The model's enhanced cybersecurity capabilities were discovered unintentionally, as they progressed faster than anticipated during the scaling process. Z.ai has implemented controls to manage these advanced features, including a \"trusted access\" approach for sensitive functionality. GLM-5.3 showcases notable advancements in coding benchmarks, with significant improvements in Terminal-Bench 3.0, DeepSWE v1.1, and AutomationBench. While not surpassing all competitors, Z.ai emphasizes efficiency, stating that GLM-5.3 completes tasks at lower token consumption compared to other models. The startup also disclosed that GLM-5.3 has identified 2,436 vulnerabilities across 269 projects, with 1,097 classified as critical or high severity. This dual focus on coding and cybersecurity raises concerns about the potential for increased offensive capabilities, as advanced models become more adept at both software engineering and security research.",
  "summary": "Chinese AI startup Z.ai, known internationally for its growing lineup of powerful, largely open source GLM series of language models, today released GLM-5.3 with substantial gains in long-horizon coding and a more consequential — and potentially sensitive — jump in cybersecurity capabilities. Already, GLM-5.3's cyber capabilities have found a \"potentially serious vulnerability in Cursor,\" the AI…",
  "key_points": [
    "Z.ai unveiled GLM-5.3 with advanced coding and cybersecurity capabilities",
    "GLM-5.3 discovered 2,436 vulnerabilities across 269 projects",
    "Trusted access controls manage GLM-5.3's enhanced cybersecurity features"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}