In the span of three weeks, our institute's worst fears have come to pass: the two top frontier labs have restricted access to their best models. None of this happened under a law, a published standard, or a process anyone can appeal. This sets an incredibly dangerous precedent.
To be clear, we believe that frontier models do indeed pose cybersecurity and biosecurity risks. But these risks have to be balanced against the risk of concentration of power. An aligned AI is nonetheless dangerous in the hands of a closed, opaque group of actors that can disable guardrails at any moment.
We have published a list of what topics you, as an AI researcher, can work on to stop the increasing concentration of AI capabilities:
https://decentralized-ai.org/research-areas
AI centralization can be countered. It's just a matter of finding the right breakthroughs.
Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.
https://openai.com/index/previewing-gpt-5-6-sol/
