Not even Microsoft is confident that Copilot is right for coding, as it’s encouraging staff to try out Anthropic’s AI model, according to one report

Even the most casual of PC users won’t have missed that Microsoft is very keen for Windows and 365 customers to use Copilot. However, according to one report, it might not actually be Microsoft’s own AI model behind the scenes, as a key internal division is being encouraged to road test the competition.

That’s according to The Verge, which claims that sources have told them staff are using Anthropic’s Claude, alongside Copilot, for AI coding tasks (aka vibe coding). Not that they’re making the choice subversively; in fact, the software giant has been happy for its staff to try out various AI coding tools, including its own GitHub Copilot, Anthropic’s Claude 4, and OpenAI’s GPT-5.

Apparently, Microsoft started the whole experiment last year, but just in the developer division, resulting in Claude Sonnet 4 being selected as the AI choice for paid users of GitHub Copilot, though the option to use others is still available. The Verge writes that now any staff member involved in ‘Experiences + Devices’ (i.e. Windows, Bing, 365, Surface, and so on) is being encouraged to experiment with Claude, even if they have no coding experience at all.

The expectation of those who can and do code, however, is that they will ultimately give feedback as to whether Claude or Copilot is the best to use, and this data will eventually shape Microsoft’s direction on what AI model is behind Copilot.

Not that this should come as much of a surprise, as many companies will routinely test what the competition has to offer, to see what can be improved in their own products. It is a bit different in this particular case, though, because Microsoft has been banging its Copilot drum so hard of late, you’d be forgiven for thinking that it’s the only AI tool worth using.

Microsoft does have more than just its staff’s preferences at stake here. In November last year, Anthropic, Microsoft, and Nvidia announced a strategic partnership, with Redmond stumping up a $5 billion investment in Anthropic, as well as the AI company committing to purchasing a cool $30 billion worth of Microsoft’s Azure compute services.

And Microsoft is being quite sensible, too, because if Anthropic’s engineers can produce a better model for coding than its own team, then it would make sense to switch to Claude.

But where does this leave OpenAI? It’s also received large investments from Microsoft and works very closely with the software behemoth. According to The Verge, that relationship remains unchanged: “OpenAI continues to be our primary partner and model provider on frontier models, and we remain committed to our long-term partnership.”

With hundreds of billions of dollars being bandied around in the world of AI, I should imagine that any such relationship is only as strong as the ultimate business bottom line: profits. Should it come to pass that future iterations of GPT, as well as GitHub Copilot, prove to be less capable at coding than Claude, it’s not hard to see where the money will eventually head towards.

That said, the most important question here isn’t about money or strategic partnerships, it’s about the future of software engineering at Microsoft and the wider world. If all things Microsoft are going to be vibe-coded by non-experts or handled by AI agents, where does that leave professional coders?

Answers on a postcard, please.