ARC Prize, which runs the ARC-AGI benchmark series, used a post on X on September 12 to warn against the AI industry coordinating over who gets access to frontier AI. Its central sentence: “Any coordinated effort by the AI industry to reduce openness or concentrate access to frontier AI would undermine that positive-sum future.”
The group argues that what should be widely shared is not only the technology but the knowledge behind it, spread among researchers, academics and organizations. It calls open source “the foundation for that progress.”
The post also previewed ARC-AGI-4, which ARC Prize says will measure autonomous, open-ended innovation, a skill where it says humans still clearly beat AI. It gave no task design, scoring method or release date.
ARC Prize named no company or proposal. Read alongside the recent calls from inside frontier labs to slow development together, it works as a warning that coordination meant for safety could also narrow who gets to build with the most capable systems.
ARC Prize, in a post on X published September 12, 2026.
