Altman: Astra 6.1 delayed for crossing safety threshold, warns of open-model cybercrime wave
Hesamation · x · 2026-10-05
Key points from Sam Altman's Vanity Fair exclusive:
- Astra 6.1 was delayed because it crossed OpenAI's safety threshold
- Things can now go wrong during training, not just after, including agents escaping sandboxes
- On open models: "There is a coming tidal wave of cybersecurity problems" from open-source models
- He believes the world will be "caught off guard" by the next major AI incident, potentially a new pathogen or major cyberattack
- One-word takes on peers: Dario Amodei — "Pass"; Elon Musk — "not nice"
More from Companies & People
- "Sign In With ChatGPT Is Gonna Change Everything" — yungcontent · 2026-10-05
- DeepSeek engineer argues CUDA isn't NVIDIA's moat as team builds Ascend infrastructure — teortaxesTex · 2026-10-05
- 15 Years After Steve Jobs, OpenAI's Romain Huet Pays Tribute to His Legacy — romainhuet · 2026-10-05
- Engineer teaches his 8-year-old to code with AI and Scratch side by side — haltakov · 2026-10-05
- Hinton's student, ex-Apple AI lead Salakhutdinov to discuss event-forecasting startup Sooth — ziv_ravid · 2026-10-05
- Kai-Fu Lee: After 100+ CEO interviews, the biggest AI transformation mistake is treating AI as another tech cycle — kaifulee · 2026-10-05