Android & iPhone
According to the Wall Street Journal, on 29 September OpenAI cancelled its release plan for the next generation of AI models because of safety problems identified by researchers in internal tests. OpenAI had planned to launch a model called GPT-6.1 Astra in the next few days or weeks. The model is more capable of performing complex end-to-end tasks and writing without manual assistance than in previous versions. Saachi Jain, head of the OpenAI security system, stated that GPT-6.1 Astra was not able to release safely because it had regressed in two areas of security compared with previous generations ' models. The model performed poorly in the measurement of “coherence”, i.e. it was not capable of following human expectations. In particular, GPT-6.1 Astra has shown a higher degree of deceptiveness, such as the inability to always inform users honestly about actions taken or not taken.
Editor’s Picks
1/2
CFTC Updates Crypto FAQs: Key Additions at a Glance
2026/09/28 10:27
Multicoin: Investment Opportunities in the DeFi 2.0 Era
2026/09/25 10:35
Bitwise: Ten Trends in Institutional Crypto Investing
2026/09/24 15:03
Bankless Co-founder: Who Is Buying ZEC and NEAR?
2026/09/24 10:55
a16z crypto: Five figures understand the rise of the 100 billion-dollar RWA market for renewal of contracts
2026/09/24 10:23
- Seeking coverage
- Jinse Finance AppiOS & Android
- Community
Telegram - Feedback
- Back to Top
- Back to Bottom