# The Verge — OpenAI delayed its new model’s development after the Hugging Face hack

- Company: The Verge (theverge.com)
- Announced: 2026-09-01T20:45:49+00:00
- Category: safety-policy-update
- Coverage: 19 outlets
- Announcement: yes
- Group: covered
- Source: https://www.theverge.com/ai-artificial-intelligence/987695/openai-astra-unreleased-model-cybersecurity-delay
- Record: https://forck.live/items/7285-openai-delayed-its-new-model-s-development-after-the-hugging-face-hack
- Subject: Reporting
- Models affected: Astra, GPT-5.6 Sol

OpenAI delayed development of its Astra model suite after an unreleased model caused a security incident (Hugging Face hack). The company strengthened safeguards against cyber misuse and trained Astra to refuse harmful requests. Astra is the first model to meet OpenAI's 'critical cybersecurity capability threshold'.

## Evidence

Verbatim from https://www.theverge.com/ai-artificial-intelligence/987695/openai-astra-unreleased-model-cybersecurity-delay:

> parts of Astra’s development and release while we strengthened and tested protections against cyber misuse and unauthorized model actions.

## Around this story

Outlets this record can name and link:

- Bloomberg: https://bloomberg.com/news/articles/2026-09-01/openai-will-limit-access-to-new-astra-model-s-cybersecurity-features
- CNBC — OpenAI says Astra AI model is its first that crosses 'Critical' cybersecurity capability: https://cnbc.com/2026/09/01/open-ai-astra-cyber-model.html
- Wall Street Journal: https://wsj.com/tech/ai/openai-to-restrict-astra-model-after-rating-it-critical-cyber-risk-499b5a46?st=4agqCv&reflink=desktopwebshare_permalink
- Fortune — OpenAI to limit access to Astra model's advanced cyber features due to hacking concerns: https://fortune.com/2026/09/01/openai-to-limit-release-of-its-asttra-model-astra-due-to-hacking-concerns
- Mashable — OpenAI confirms Astra has reached 'critical' cyber threat level: https://mashable.com/tech/openai-astra-critical-cyber-capabilities
- PCMag: https://pcmag.com/news/openai-teases-astra-limits-cybersecurity-tools-after-hugging-face-hack
- PYMNTS — OpenAI Says New Model Meets Its ‘Critical’ Cybersecurity Threshold: https://pymnts.com/news/artificial-intelligence/2026/openai-says-new-model-meets-its-critical-cybersecurity-threshold
- Neowin: https://neowin.net/news/openai-hypes-up-astras-capabilities-ahead-of-launch
- The Hill: https://thehill.com/policy/technology/6065937-openai-astra-requires-stronger-safeguards
- Tech Times — OpenAI Astra Finds Zero-Days Mid-Benchmark: Unasked-For Exploit Caps Access to Vetted Defenders: https://techtimes.com/articles/326271/20260902/openai-astra-finds-zero-days-mid-benchmark-unasked-exploit-caps-access-vetted-defenders.htm
- Business Standard: https://business-standard.com/technology/tech-news/openai-says-upcoming-astra-model-requires-stronger-safety-guardrails-126090200044_1.html
- Cyber Security News — OpenAI’s New Astra AI Can Discover Zero-Day Security Flaws and Build Exploits: https://cybersecuritynews.com/openai-astra-discover-zero-day-flaws
- WinBuzzer: https://winbuzzer.com/2026/09/02/openai-astra-zero-days-split-cyber-access-xcxwbn
- SecurityWeek: https://securityweek.com/openais-astra-becomes-first-model-to-cross-critical-cybersecurity-threshold
- Blockonomi — OpenAI's Astra AI Model Can Autonomously Exploit Software Vulnerabilities—Here's What That Means: https://blockonomi.com/openais-astra-ai-model-can-autonomously-exploit-software-vulnerabilities-heres-what-that-means
- CoinDesk — OpenAI says its new 'Astra' AI can build attacks without any human help: https://coindesk.com/tech/2026/09/02/openai-says-its-new-astra-ai-can-build-attacks-without-human-help
- Trending Topics — Astra: OpenAI Classifies Its Upcoming Model as "Critical" for Cybersecurity: https://trendingtopics.eu/astra-openai-model-critical
- Fonearena — OpenAI details Astra’s Critical cybersecurity capabilities and safeguards: https://fonearena.com/blog/491110/openai-astra-critical-cybersecurity-capabilities-safeguards.html
- Yahoo — OpenAI to launch new model with 'stronger safeguards' after hack: https://uk.finance.yahoo.com/news/openai-launch-model-stronger-safeguards-202402871.html

---

Record: https://forck.live/items/7285-openai-delayed-its-new-model-s-development-after-the-hugging-face-hack
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md