Disney drops GitHub Copilot for OpenAI Codex — and keeps Claude and Cursor, not Amazon's tools
Disney is dropping GitHub Copilot and Amazon's Kiro and Q for US engineering staff this August, replacing them with OpenAI's Codex while keeping Anthropic's Claude Enterprise and Cursor — a real-world vote on which AI coding tools survive enterprise scrutiny.
20 August 2026
Disney is dropping GitHub Copilot and Amazon’s Kiro and Q for US engineering teams this August, according to an internal message reported by Business Insider. In their place: OpenAI’s Codex, alongside Anthropic’s Claude Enterprise and Cursor, which Disney is keeping. Neither Disney nor Microsoft has confirmed it publicly, but the reported reasoning is blunt — engineers said they rarely used Copilot, and one product manager described its output as “needlessly complex.”
This lands weeks after GitHub moved Copilot to usage-based billing, and it’s a useful data point precisely because it’s not a survey or a vendor’s own benchmark — it’s one of the world’s largest media companies making a procurement call with real budget behind it. Three tools kept, two dropped, out of five. Disney’s list matches what shows up repeatedly in independent testing this year: Claude Code and Cursor scoring at the top for real engineering work, Codex gaining fast among teams already inside the OpenAI ecosystem, and Copilot losing ground now that it’s competing purely on capability rather than default IDE placement.
The signal for anyone commissioning software isn’t “pick these three tools.” It’s that AI coding tool choice has stopped being a foregone conclusion tied to whichever platform you’re already on. Enterprises are now auditing which agents actually produce usable code and cutting the rest — including tools from vendors they have broader relationships with, like Amazon.
So what
If a development partner tells you their AI tooling stack is fixed rather than actively evaluated, that’s worth a follow-up question — the tools worth using in August aren’t the ones that were default six months ago. Our AI-assisted development work runs on whichever agent fits the job at hand, reviewed against real output rather than brand loyalty. Get in touch if you want a second opinion on what your current build partner is using and why.