Search

Why not Sovereign AI?

Jul 30
2 min read

Written by Swagat Irsale, Growth Advocate.

I am Swagat Irsale and I am a Growth Advocate who consults with start ups and scale ups in growing revenue and scaling products.

--------------------------

1 - A month back one of the AI companies announced, we have created something so good that we are scared to release it. I am impressed. It caught government attention as well.

Fire, locomotive, refrigeration, computer, internet/ mobile and now AI.

Humanity is getting better sisters and brothers … 🙋.

2 - Recently another AI company said - our AI mistakenly penetrated Hugging Face.

Think about it; to do this you have to target a web property, have a plan, escape from a sandbox environment, prepare to infiltrate and execute the plan.

And then teams from both companies work together for 24 hours and confess that no major damage has been done and the intent was not malicious.

Please note that, penetrate, infiltrate (similar) are the words used in the media and not by me.

Considering this is not a marketing gimmick (?) and if AI really did all this targeting, planning and execution then I feel we are close to AGI in the next couple of years.

AGI, sisters and brothers; I am excited as a consumer … 👍.

3 - Try this to test the ground. Ask Gemini (and or its peers) a plan, give some context, prompt, etc. By default it takes information from Gmail, meet, Docs, Drive, etc.

Recently Gemini gave me feedback regarding one of my clients. I did not share any client details with Gemini. Sometimes it takes partners such as Zoho, Zoom Info, etc. as clients as their information exists in emails/ meetings; we all know it halu-s.

Here is the point = AI Output is so so as of today if I iterate multiple times.

40% of the AI output is usable though ... 🏆.

AI is innocent like a child in 2026. We are feeding it a lot of context and it is collecting information from many of our personal, work sources.

I remember a song; will mention at the end of the post. Kya se Kya (you get the hint) …

Here is what I plan to do?

AI companies are striving for global attention (competition, ROI on AI, open source, etc.?). There may not be boundaries on how far these stunts (?) can go.
I am thinking of doing multi-homing to keep my data a bit private. So may be take calls on Zoom instead of Google meet; citing my above experience.
The Web, its components and the previous era were not intelligent/ autonomous, but AI is getting intelligent day by day. Maybe I should learn to protect my intellectual property (IP) from leakage.