Q&A
Discussion of AI Action Plan components and risk assessment process
1:02:25
·
3 min
Council Member Jennifer Gutiérrez inquires about the specifics of the AI Action Plan, focusing on the AI risk assessment process that is currently in progress. Alex Foard from the Office of Technology and Innovation (OTI) explains the high-level steps involved in creating a risk landscape and interfacing with existing frameworks.
- OTI is working on creating an exhaustive analysis of potential AI risks
- They are examining how AI risks interface with existing privacy and cybersecurity frameworks
- The AI steering committee, composed of diverse city agencies, will help inform policy development
- The process aims to operationalize abstract concepts like AI bias into actionable steps
Jennifer Gutiérrez
1:02:25
Okay.
1:02:26
I wanted to ask a little bit more about that action plan.
1:02:28
Now just some of the the like, the specifics because I think they're and and, specifically, they're I know there's some are in progress, some are complete.
1:02:39
Right?
1:02:39
And so I just wanna get a little bit more as much as you can share about some of the ones that are in progress, I know that in 1.6, an AI risk assessment and 1.8 monitoring tools are the most important are the most important, and, obviously, it's still in progress.
1:02:58
Is there anything more that you can tell us about risk assessment and monitoring not in the action plan right now?
Alex Foard
1:03:05
So yeah.
1:03:05
I mean, I can certainly speak to some of the high level steps that I think we think are really important here.
1:03:10
So one of those, of course, is is sort of creating that landscape of what we think their possible risks could be.
1:03:17
Mhmm.
1:03:18
We think what is really important is to to really do as exhaustive in analysis as possible to say, here's the different ways that risk could rise.
1:03:28
And then subsequent to that to say, okay.
1:03:30
Now that we know what these categories of risk risk are, are there any existing frameworks that already address these?
1:03:35
Right?
1:03:36
So we've talked about privacy and cybersecurity.
1:03:38
That's a place where we wanna say, okay.
1:03:40
We've got the office of information privacy.
1:03:42
We have the identifying information law.
1:03:44
We have our office of Cyber Command.
1:03:46
Before we sort of approach addressing AI risk for information privacy or cybersecurity, in a vacuum, let's see how it interfaces some of these other processes and frameworks.
1:03:57
And then from there, we can say, okay.
1:03:58
What is not being counted for?
1:04:00
What are the gaps that we're trying to fill?
1:04:01
And how do we see this as a process that helps us turn something that can sometimes do a bit of an abstract concept into something that makes sense for us in an agency.
Jennifer Gutiérrez
1:04:11
And and I'm sorry.
1:04:12
What's the the last step?
Alex Foard
1:04:14
Of being able to sort of, like, operationalize that as a process.
1:04:17
Right?
1:04:17
So how we say, you know, this is a complex abstract idea, you know, things like bias and AI, Mhmm.
Darío Maestro
1:04:23
We
Alex Foard
1:04:23
wanna be able to sort of say how can we actually make that a a real thing that we can address versus just talking about in the abstract.
Jennifer Gutiérrez
1:04:30
And are there specific agencies that you either a a part of the steering committee or not that you are looking to be able to, like, kind of prioritize or spend more time on assessing gaps that need to be filled, opportunity for operationalizing.
Alex Foard
1:04:48
Well, certainly, the steering committee will be a valuable resource for us.
1:04:50
That's why the the steering committee exists to be able to help inform that policy development.
Jennifer Gutiérrez
1:04:54
And I'm sorry to interrupt.
1:04:55
Can I just and can I also ask the agencies that are in the steering committee?
1:04:59
Did they they opted in to being?
1:05:01
Like, did they did they wanted like, did you all do the reach like, the outreach
Alex Foard
1:05:05
or the outreach?
Jennifer Gutiérrez
1:05:05
You did the outreach.
1:05:06
Okay.
1:05:07
Okay.
1:05:07
Great.
1:05:07
So the assumption is that they're, like, they're part of it.
1:05:10
They're also looking to maybe it kind of integrate or, like, activate their agencies along the action plan a little bit.
Alex Foard
1:05:19
So the goal with the composition of the steering committee was to get as much diversity in terms of the sort of mission that different agencies have.
1:05:26
We have agencies of different sizes to reflect different sort of magnitude of technology, different perspectives on the components of AI that they care about most.
1:05:36
Some of them are are operational agencies.
1:05:38
Some of them are policy agencies.
1:05:39
So we wanted to have a very diverse collection there.
1:05:43
And the general goal of the steering committee is to help inform that policy development.
1:05:47
Right?
1:05:47
So they can better inform us This is what this would look like on the ground as an operational agency.
1:05:53
These are some of the other community groups that we care about, etcetera.
1:05:56
So they're there to inform that work.
