Closed DelaramGlp closed 5 months ago
Hi. Thanks for the review.
- Do we want to add sources for the AI Act concepts, i.e. article numbers, or shall we wait for the official publication?
Let's wait for official publication and article numbers.
2. Definitions needed for `eu-aiact:FRIA` and `eu-aiact:HighRiskAIAssessment`.
Added: FRIA - An assessment undertaken to evaluate how the system might impact fundamental rights, and High Risk AI Assessment - An assessment undertaken to determine whether the AI system is classified as high-risk (we will replace with official descriptions from the final version?)
3. `eu-aiact:AIOperator` should be added as a parent for `eu-aiact:AIProvider`, `eu-aiact:AIDeployer`, `eu-aiact:AIDistributor`, `eu-aiact:AIImporter`, `eu-aiact:AuthorisedRepresentative`.
done
I think
eu-aiact:ProductManufacturer
needs to be added, though there is no definition for it. Okay, added asAIProductManufacturer
for consistency with other terms, and with definition as "entity that manufactures the product" and parent astech:Manufacturer
The link to
pd:Biometric
is missing. Where? Do you mean in the related field for term? If so, then it will be fixed with #161See more section for all the capabilities needs to be fixed "section [PURPOSE] [CAPABILITIES] (https://w3c.github.io/dpv/legal/eu/aiact/#vocab-purpose)". The links goes to the capability section but you might want to change the URL to https://w3c.github.io/dpv/legal/eu/aiact/#vocab-capability.
Also, in the introduction taxonomy of purposes should be replaced with taxonomy of capabilities. Thanks, this is probably left over from when we had 'purpose'. I've changed the module name to capability now.
eu-aiaact:BiometricIdentityVerification
: is it a Purpose or Capability? In DPV identity verification is a purpose, so this would be a purpose here as well? It should have been a capability but that would mean going back and changing the DPV purposes list as well. For the AI Act, I'm fine if this is considered a capability for consistency for now.
eu-aiaact:EmotionRecognitionSystem
: related termsaia:EmotionRecognition
should beeu-aiact:EmotionRecognition
fixedBased on the definition of
eu-aiact:NationalCompetentAuthority
, it is parent ofeu-aiact:NotifyingAuthority
andeu-aiact:MarketSurveillanceAuthority
. added
dpv:risk
needs to be fixed:dpv:Risk
with the following link: https://w3c.github.io/dpv/dpv/#Risk Where? I found one in eu-aiact:Risk definition, fixed that.
risk:Misuse
is missing in the risk extension. What would this be? A consequence? (I've added it to consequence for now)
Implemented all of the above. See:
Thanks.
To me Misuse seems like a risk or risk source, but I'm not sure.
Shall we add Prohibited AI Assessment as well? Also we need a concept for Risk Management System.
For the next release, we need to add risk management concepts from Art. 9 and link them to ISO 31000 concepts (#74)
Misuse can indeed be a risk/source - but we'll get to that when go through the entire taxonomy in next iteration. There can be too much overlap between consequence and risk source, so we should try to create a single taxonomy and let the use-case specify it as a risk source or consequence where possible (or declare it as both source and consequence). I don't think we will have a taxonomy of risks - because then how to distinguish between source, risk, and consequence - concepts can be any of those depending on the use-case.
Prohibited AI Assessment - your call.
Risk Management System - for AI Act or in RISK extension? I think it is a type of technology for performing in Risk Management (which should be in RISK extension)?
Agree, risk management system should go into the RISK extension.
If you have a definition at hand - I can add it right away.
This is from the AI Act, Art. 9: The risk management system shall be understood as a continuous iterative process planned and run throughout the entire lifecycle of a high-risk AI system, requiring regular systematic review and updating.
This is from ISO 31000, 3.2: risk management: coordinated activities to direct and control an organization with regard to risk.
Looking into these definition, should risk management be added to RISK and risk management system to AI Act?
From ISO 31073: risk management - coordinated activities to direct and control an organization with regard to risk; and risk management process - systematic application of management policies, procedures and practices to the activities of communicating, consulting, establishing the context, and identifying, analysing, evaluating, treating, monitoring and reviewing risk. I found this in ISO 42001: management system: set of interrelated or interacting elements of an organization to establish policies and objectives, as well as processes to achieve those objectives.
So seems to be pedantic differences in terms, e.g. to refer to the 'system' and the 'process'. We can put risk:RiskManagementProcess
as subclass of risk:RiskManagement
with a note stating that it is similar to 'risk management system' and using the 42001 definition, and then have eu-aiact:RiskManagementSystem
be a subclass of risk:RiskManagementProcess
. What do you think?
Makes sense.
I added in risk:RiskManagement
as the broad concept and didn't include RiskManagementProcess
as I couldn't find a source for the term. eu-aiact:RiskManagementSystem
as a subclass of risk:RiskManagement
is clear and consistent with use of standards with the law. (reopen the issue if more changes are to be made)
Do we want to add sources for the AI Act concepts, i.e. article numbers, or shall we wait for the official publication?
Definitions needed for
eu-aiact:FRIA
andeu-aiact:HighRiskAIAssessment
.eu-aiact:AIOperator
should be added as a parent foreu-aiact:AIProvider
,eu-aiact:AIDeployer
,eu-aiact:AIDistributor
,eu-aiact:AIImporter
,eu-aiact:AuthorisedRepresentative
.I think
eu-aiact:ProductManufacturer
needs to be added, though there is no definition for it.The link to
pd:Biometric
is missing.See more section for all the capabilities needs to be fixed "section
[PURPOSE][CAPABILITIES] (https://w3c.github.io/dpv/legal/eu/aiact/#vocab-purpose)". The links goes to the capability section but you might want to change the URL to https://w3c.github.io/dpv/legal/eu/aiact/#vocab-capability. Also, in the introduction taxonomy of purposes should be replaced with taxonomy of capabilities.eu-aiaact:BiometricIdentityVerification
: is it a Purpose or Capability?eu-aiaact:EmotionRecognitionSystem
: related termsaia:EmotionRecognition
should beeu-aiact:EmotionRecognition
Based on the definition of
eu-aiact:NationalCompetentAuthority
, it is parent ofeu-aiact:NotifyingAuthority
andeu-aiact:MarketSurveillanceAuthority
.dpv:risk
needs to be fixed:dpv:Risk
with the following link: https://w3c.github.io/dpv/dpv/#Riskrisk:Misuse
is missing in the risk extension.