GitHub Copilot Is Switching to Pay‑Per‑Token

GitHub Copilot Is Switching to Pay‑Per‑Token

Starting June 1, 2026, GitHub Copilot will stop charging a flat monthly rate and start billing users based on how many tokens they use — basically, how much text you type in and how much Copilot writes back.

Here’s what that means in plain English:

  • Old system: You made a set number of “requests” per month, no matter how big or small.
  • New system: Every bit of text — your inputs, Copilot’s outputs — is counted and billed by tokens (a token is roughly three‑quarters of a word).

Pricing details:

  • Copilot Pro still costs $10/month, but now you get 1,000 “AI credits” instead of unlimited requests.
  • Simple questions will barely use up any credits. Huge coding tasks? Those will burn through your balance fast.
  • Fancy new models cost more tokens than basic ones.
  • Autocomplete and Next Edit suggestions remain free (for now, at least).

GitHub (and parent company Microsoft) say the change follows what OpenAI and Anthropic are already doing. In short: the AI world has gone from subscriptions to micro‑transactions.

So, come June, keep an eye on your token count — or risk your Copilot running out of fuel mid‑debug.


SAP Says: Govern Your AI or Get Governed by It

SAP’s Manos Raptopoulos has a message for big business: your AI isn’t a cute little assistant anymore—it’s basically an employee with infinite coffee and no common sense. And if you don’t manage it properly, it’ll wreck your operations faster than you can say “rogue agent.”

In corporate-speak, this means: set strict rules, track every digital sneeze, and never let your model “just try things.” Because in enterprise AI, 90% accuracy = disaster, and 100% = profit.

He warns that agentic AIs—those clever bots that plan, reason, and act on their own—could cause an “agent sprawl” worse than that old shadow IT mess, only more expensive. The cure? Governance, governance, and—oh yes—more governance.

The secret sauce, according to SAP:

  • Ground AI in real company data, not random internet vibes.
  • Build role-based AI personas (your CFO gets one that doesn’t daydream about poetry).
  • Make interfaces intent-based, so employees can just say what they want and the system does it—without catching fire.

In short: good data, good governance, and realistic ambitions will make AI your best employee. Skip any of those and your “smart assistant” becomes the world’s costliest intern.


*Disclaimer: The content in this newsletter is for informational purposes only. We do not provide medical, legal, investment, or professional advice. While we do our best to ensure accuracy, some details may evolve over time or be based on third-party sources. Always do your own research and consult professionals before making decisions based on what you read here.