Before You "Rage-Report" a GenAI Issue

Before You "Rage-Report" a GenAI Issue

I'll admit, I've impatiently sent feedback when using GenAI tools. When a hallucination happens, the QA people at OpenAI, Google/Gemini, and Microsoft (Copilot) have to be told, right?

If you’re the kind of person that likes to help improve software, or if you just can't let it slide, you can provide feedback to people working at Microsoft. It’s one of the ways they legitimately fine tune their algorithms, but it has risks.

So.... be advised if you leave the default clicked to "Yes" in Copilot for Microsoft 365 (below).

Copilot asks if it’s OK to “My prompt and generated response" which are usually NOT seen by any person working at Microsoft, due to the data protections in place. Furthermore, "relevant content samples, and additional log files” will be sent.

What's that mean?

When using Word, a “contextData.json” link appears which, if clicked, shows the exact prompt and response that will be sent (as shown). It’s OK to click that hyperlink (and actually quite interesting) to see how your prompt is augmented before it’s sent to the LLM.

But……

If you click “Yes” here, the prompt and any document or spreadsheet that was referenced in the prompt will be sent to Microsoft. See more detail at https://support.microsoft.com/en-us/topic/providing-feedback-about-microsoft-copilot-with-microsoft-365-apps-c481c26a-e01a-4be3-bdd0-aee0b0b2a423

There’s a chance a person at Microsoft or OpenAI will look at that data. So, to be safe, ask your management or IT leaders if you should include the prompt and content samples. Unless your company allows it, you should not click “Yes” to share this data with Microsoft, especially if the .doc or .xls you referenced has organizational information in it.

IT Pros and CISOs can configure specific feedback controls. See more detail at https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-feedback-ms-org?view=o365-worldwide#specific-policies-you-can-configure

Moreover, this kind of do/don't decision should be outlined in a GenAI Policy published by the organization's CISO or AI Center of Excellence. See more detail about Preparing for Copilot for Microsoft 365 - eGroup Enabling Technologies (egroup-us.com)


要查看或添加评论,请登录

Chris Stegh的更多文章

  • The Sweet Spots for Microsoft 365 Copilot

    The Sweet Spots for Microsoft 365 Copilot

    Where It Shines and Where It Struggles Not all tasks are created equal when it comes to AI automation. This blog offers…

  • Using DeepSeek to Improve AI Governance

    Using DeepSeek to Improve AI Governance

    DeepSeek's media blitz is a wake up call for US AI companies, especially those whose valuations are built on frontier…

    2 条评论
  • Creating a SharePoint Agent was Easy. Perfecting it was harder!

    Creating a SharePoint Agent was Easy. Perfecting it was harder!

    It's no secret that Large Language Models (LLMs) can create answers that miss the mark. “Confabulations,” or worse…

    1 条评论
  • ?? The Future of Robotics in Our Homes: Vision vs Reality

    ?? The Future of Robotics in Our Homes: Vision vs Reality

    Why don’t we have C-3PO-like droids in our homes yet? Hal was supposed to be here in 2001! The arrival of GenAI and…

  • A Maturity Model for Collaboration

    A Maturity Model for Collaboration

    The Impetus for a Model There’s no shortage of productivity tools, but there is little means of judging their efficacy.…

  • Copilot Coming to Microsoft 365

    Copilot Coming to Microsoft 365

    You've by now seen the previews of Microsoft’s unique capabilities using large language models within Microsoft 365…

    1 条评论
  • A Model for Collaboration Maturity

    A Model for Collaboration Maturity

    The Impetus for a Model There’s no shortage of productivity tools, but there is little means of judging their efficacy.…

  • Back with People! RSA Conference Summary

    Back with People! RSA Conference Summary

    RSA Conference 2022 wrapped last night, and after a week of absorbtion, it's time for some emission. The FUD was…

    1 条评论
  • Clock starting for End of Life of Legacy Auth to Exchange... Again!

    Clock starting for End of Life of Legacy Auth to Exchange... Again!

    For good reason, MSFT has twice extended the end of life of legacy authentication protocols hitting Exchange Online…

  • Preparing for Human-Operated Ransomware

    Preparing for Human-Operated Ransomware

    Ransomware is now a (White) household name. How serious has it gotten? Critical energy, food, transportation, and…