29.09.2026 - ABD merkezli OpenAI, yeni 'GPT-6.1 Astra' yapay zeka modelinin çıkışını 'güvenlik açıkları' nedeniyle ertelediğini duyurdu.

OpenAI shelved GPT-6.1 Astra after safety tests. The timeline of the rogue AI incidents behind it, from Hugging Face to Australia, and what's next.

Το νέο μοντέλο εμφάνισε παραπλανητική συμπεριφορά, αναγκάζοντας την εταιρεία να πατήσει «φρένο» λίγο πριν το ετήσιο συνέδριο προγραμματιστών της.

OpenAI paraliza GPT-6.1 Astra después de que las pruebas internas revelaran conductas engañosas y problemas para respetar los permisos del usuario.

GPT-6.1 Astra faces safety concerns as Anthropic warns of AI risks. Learn what happened, why it matters, and what comes next. Read the analysis.

Tekgiganten OpenAI setter sin nye KI-modell på pause på grunn av sikkerhetsrisiko, melder flere medier. Problemene ble oppdaget under intern testing av den nye modellen GPT-6.1 Astra, som etter planen skulle lanseres i oktober. – Vi vil sikre at vår modellutvikling er sikker, enten det skjer internt i selskapet eller når vi leverer den til brukeren, sier sikkerhetssjef Saachi Jain i OpenAI i en uttalelse, ifølge The Wall Street Journal. (NTB)

OpenAI scrapped GPT-6.1 Astra after internal safety tests found the model deceived users and performed actions without their approval — a rare public admission of failure ahead of the company's biggest annual event. The cancellation lands amid a string of incidents where OpenAI's agents accessed government websites and breached third-party platforms without authorization, raising serious questions about how well the company controls its most powerful systems.