They did...
They're literally just passing on the costs https://platform.claude.com/docs/en/about-claude/pricing
Anthropic just provides a subscription - which Enterprise usually doesn't want you to use because everything you're submitting through that will be trained on / becomes part of their model.
So If you use it without explicit permission from your employer you may be committing a contract violation which can have serious consequences - up to jail time - as they can sue you for that.
My Pro account very clearly has a toggle for "Help improve our AI models: Allow the use of your chats and coding sessions to train and improve Anthropic AI models."
> Our use of Materials [...] Even if you opt out, we will use Materials for model training when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance our safety research.
The last part is essentially a catch all, which let's them train on everything they want - and they probably are.
But the important bit here isn't actually wherever they're actually training on it - that doesn't matter from the legality aspect of it. You're liable anyway, as all contracts I've ever signed explicitly forbid me from sharing internal data of any kind (including code) with third parties.
You can be prosecuted just from using it - wherever anthropic decides to train it's model on it or not.
If you use Claude via API in your own app, you're paying full price.
If you have an "API Plan" for Claude Code (i.e., free), you're paying full price.
If you have a Pro, Max, Max 5x, or Max 20x, your tokens are subsidized up until a rate limit. Then you pay full price for usage thereafter, until the end of the billing cycle.
The widespread belief in industry right now is that the per-seat pricing (which Copilot bailed from first) is going to go away in the near-term.
At my company we did the comparison and Copilot still wins: for 20$ you get a seat and 20$ of usage, whereas with Claude enterprise you get a seat and then usage is completely added. Moreover usage in Copilot is exactly the price of the providers AND it allows us to use various models from multiple providers.
The case that might be less expensive is if you negociate a volume discount with AWS for Bedrock usage, but that is also possible with GitHub and Microsoft.
Edit: wording on the cost saving effort
I've swapped to the 20x Claude plan for a month or two to knock out two ideas I need to get it MVP - expecting Claude to go token priced soon.
pleasantly surprised, claude's way ahead in tooling but the ability to designate what model your subagents use and having access to all models is a better feature than all of what claude offers combine atm.
The only limit on the amount of ai can consume in a month a work is dollars, so anything that helps with cost is the best model/harness for me.
It also did a better job at smart designating subagents itself where as claude often used higher cost models.
For example: https://github.com/monooso/dotfiles/tree/main/.claude/agents
Honest question, can you ellaborate? If given the option, I use OpenCode but what do you find in Copilot CLI that makes you prefer it to Claude Code?
There is also IMO a distinct difference in "tone" in the dialogue. Claude seems to impersonate a human a bit more than I like.
Claude is of course very good as well and does a few things better than copilot too, but overall I'd prefer to use Copilot.
For my personal work, I still use Claude Code as its cheaper and the limits don't bother me to much, but it feels a bit like being handcuffed to Anthropic vs being at work and freely selecting models.
I also use the Copilot ACP server inside Pycharm and that works decently well too, although it has some annoying bugs, but if you're a Jetbrains user you're used to annoying bugs.