Summary: OpenAI’s Rest-of-World Terms of Use took effect on January 1, 2026 for users outside the US, UK, and EU. The same opening period included GPT-5.2’s December release and reporting that Meta’s Avocado model may ship proprietary rather than open source.
New Year’s Day 2026 brought OpenAI’s updated Rest-of-World Terms of Use into force. The policy page spells out usage rules for everyone outside the US, UK, and EU — meaning most of the world’s markets now operate from a single rulebook, one deliberately kept apart from the three largest.
Two more signals belong on the same timeline. OpenAI’s model release notes show the GPT-5.2 base model shipping on December 11, 2025. And on December 9, 2025, CNBC reported that Meta is building a frontier model codenamed Avocado that may arrive as proprietary rather than open source — a clear departure from its Llama-era playbook. Terms, model cadence, and open-source strategy converged inside the year’s first week.
Tighter Terms: Geography Becomes a Product Boundary
That a Rest-of-World policy exists at all is the loudest message. When a company fences off “everywhere except the US, UK, and EU” as its own jurisdiction, regulatory divergence has burst the seams of any single global document. Teams embedding OpenAI products feel the practical effects fast:
- Rights and obligations for users across covered markets now flow from this one document, so multi-market products owe themselves a compatibility review
- Once terms are regionalized, feature rollouts and data policies gain permission to diverge by region as well
- Legal review couples tightly to the product roadmap; a terms change now demands product-change handling
Model Cadence: The Acceleration After GPT-5.2
The GPT-5.2 base model went live December 11, 2025 — three weeks shy of the new year. Entering 2026 with the freshest base model already planted usually heralds a year of derivative releases and capability expansion built on stable ground. For API customers, the price of acceleration reads plainly: shorter effective model lifetimes, migrations demoted to routine engineering, and version pinning plus model routing abstraction promoted from optional extra to baseline practice.
Two second-order effects trail behind. First, the pricing and capability assumptions baked into last year’s products spoil sooner, so unit economics deserve review more often than annually. Second, evaluation harnesses climb in value: with the models underneath you churning, a steady benchmark suite is the only instrument that distinguishes “this upgrade genuinely helps my workload” from “this upgrade quietly regressed it.”
The Open-Source Split: Avocado’s Turn
CNBC’s reporting places a strategic question squarely on the table: Meta’s Avocado may not be open source. Should that hold, the Llama-era trade of open weights for ecosystem reach faces a serious rewrite, and the supply of frontier-grade open models thins. Teams that lean on open weights should log two consequences:
- The pool of self-hostable frontier models contracts, tipping bargaining power toward closed APIs
- Model sourcing needs open and closed tracks running in parallel, so a wrong bet on either side never strands the product
What the Three Threads Share
Laid side by side, the opening weeks of 2026 aim at one direction: commercial and legal geography is fragmenting, model iteration has shifted into high gear, and open source no longer serves as the industry’s default route. The pragmatic answer for developers and product teams is to treat change itself as a design input — assume terms, model versions, and licensing regimes all move during the year, and build architectures with substitution room baked in, rather than absorbing each shift as an annual ambush.
For Meta, the interesting question points the other way. If Avocado ships proprietary, the company is wagering that frontier capability outweighs the ecosystem goodwill open weights generate — a wager the market will grade before the year is out.