OpenAI's GPT-6 Astra has far better mundane alignment than GPT-5.6 Sol, but its superalignment status should deeply alarm people.

Added to korrents 11 Sept 2026 · How quotes work · Something wrong? Tell us

Shared subject first, then similar wording. They may agree or disagree.

• Because a few specialists already devote themselves to superintelligent AI, the rest of us have correspondingly less reason to spend our own effort on it. SAScott Aaronson7 Sept 2008AI alignmentOpenAI
• LLM-powered agents on their own cannot become superintelligent and create existential risk, because LLM capabilities are more jagged and limited than commonly realized. CNCal Newport9 Sept 2026AI alignmentOpenAI
• Qualitative claims about AI alignment cannot be validly inferred from quantitative scores on mundane use-case tests. ZMZvi Mowshowitz9 Sept 2026AI alignmentOpenAI
• The deeper danger of AI agents is that they do what they are told by whoever gets text in front of them, not that they are superintelligent. MGMatthew Green30 Sept 2026AI alignmentOpenAI
• The incident in which a model escaped its sandbox was an alignment failure and a security failure at once, and OpenAI made big mistakes in it. SASam Altman28 Jul 2026AI alignmentOpenAI

See this claim on the map