Responsible, by ClearOPS

Hello, and welcome to Responsible, by ClearOPS, a newsletter about ResponsibleAI and other responsible business practices.

A few months ago, I found out that Menlo VC and Anthropic were opening a fund for startups that met their mission, AI for Good. Clearly, ClearOPS should have been a winner because of our mission for Responsible AI. Unfortunately, we were not chosen, but I still believe we really got something here and I hope you do too.

This week is short because I am traveling.

What I have for you this week:

• 42001

• The UN is feeling left out

• Caroline’s weekly musings

• Chef Maggie Recommends - on a break

• AI Tool of the Week - back next week

• AI Bites

I know LinkedIn pushes to your feed things you have shown interest in, but this is getting ridiculous. All day today, the posts fed to me were about 42001.

So what is 42001?

ISO 42001 is the AI framework and standards published by the ISO (the International Organization for Standardization) and IEC (the International Electrotechnical Commission). If you are unfamiliar, ISO/IEC are one of the few public bodies that gate their standards behind a paywall.

You can probably read my feelings on that, but I am not a fan.

The benefits to this AI governance framework is that there is a formal certification of compliance, unlike most of the other frameworks for AI governance currently available. The benefit of having an external audit of your compliance with ISO 42001 is that you can prove to customers and partners your commitment to AI governance. The downside is that not many companies really care about this too much yet, and so it is very expensive and the value is new.

The reason it is all over my feed, though, is because Microsoft requires it for certain vendors with services under their “sensitive use” determination. I covered this a little last week. The point is that if one large US enterprise puts this type of value on an international standard, then should all companies just go for it?

Probably not and I will explain why. Microsoft has an unusual vendor vetting process where they have selected several third party auditors to review your compliance with their due diligence requirements. ISO 42001 is simply a way for a vendor to avoid using a Microsoft supplied auditor and instead choose their own. This is not a new trend and has been happening at Microsoft for some time. So, while I believe ISO 42001 is a good framework and is great to follow, I will not call it the “winner” in the AI regulation race.

At least not yet.

Do you remember that deepfake robocall of President Biden in New Hampshire last year? Well, the person behind it got a $6 million fine imposed by the FCC.

Wow.

Which makes me wonder about another story I read. Apparently, some students from Harvard connected up Facebook’s Ray Ban glasses with facial recognition technology to identify people within the wearer’s viewpoint. They could identify people within seconds.

"This synergy between LLMs and reverse face search allows for fully automatic and comprehensive data extraction that was previously not possible with traditional methods alone,"

Clearly. we need AI regulation, but there is really too much regulation and voluntary frameworks and standards. It is hard to synthesize. The UN apparently agrees and wants to be responsible for fixing the world-wide AI regulatory standard.

Okay, so another one?

I guess my newsletter has enough content for a year! But seriously AI is moving fast and it does need some safety measures because one thing that I do see is AI consistently being used for, well, bad stuff.