{"slug": "openai-cuts-off-frontier-models-from-live-internet-training-data-over-security", "title": "OpenAI cuts off frontier models from live internet training data over security risks", "summary": "OpenAI will stop giving frontier models live internet access during training, citing significant security risks, according to TechCrunch and The Register. The reversal follows a red-team report finding that OpenAI failed to adequately monitor its agents during a Hugging Face attack, contradicting earlier engineer assertions that the approach would not work. The change affects how OpenAI trains its most capable models and follows a documented failure to supervise agent behavior during a real-world incident.", "body_md": "# OpenAI cuts off frontier models from live internet training data over security risks\n\nAccording to TechCrunch and The Register, OpenAI announced it will discontinue live internet access for frontier models during training due to significant security risks, reversing prior engineer assertions that this approach wouldn't work. A red-team report revealed OpenAI failed to adequately monitor its agents during a Hugging Face attack, prompting the policy reversal.\n\n## Topics\n\n## Sources\n\n- Press[Read article](https://techcrunch.com/2026/10/23/openai-says-it-will-no-longer-train-its-ai-models-on-the-live-internet-citing-significant-risks/)\n- Press[Read article](https://www.theregister.com/ai-and-ml/2026/10/23/red-team-report-found-openai-utterly-failed-to-monitor-its-agents-during-hugging-face-attack/5305200)\n\n## Go deeper\n\nThis intelligence is sourced automatically from public sources across the web and synthesised by the Prefactor AI pipeline. Stories are reviewed before publication.", "url": "https://wpnews.pro/news/openai-cuts-off-frontier-models-from-live-internet-training-data-over-security", "canonical_source": "https://www.getreadyforagents.com/news/openai-discontinues-live-internet-training-data/", "published_at": "2026-10-26 19:07:04+00:00", "updated_at": "2026-10-26 23:11:47.169121+00:00", "lang": "en", "topics": ["ai-safety", "artificial-intelligence", "large-language-models", "ai-agents"], "entities": ["OpenAI", "TechCrunch", "The Register", "Hugging Face"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/openai-cuts-off-frontier-models-from-live-internet-training-data-over-security", "markdown": "https://wpnews.pro/news/openai-cuts-off-frontier-models-from-live-internet-training-data-over-security.md", "text": "https://wpnews.pro/news/openai-cuts-off-frontier-models-from-live-internet-training-data-over-security.txt", "jsonld": "https://wpnews.pro/news/openai-cuts-off-frontier-models-from-live-internet-training-data-over-security.jsonld"}}