CarefulAI
  • Accelerating Safer AI Investments
  • Accelerating Safe AI Users
  • Accelerating Safe AI Providers
  • AI Assurance
  • Learning Academy
  • Public Interest AI
  • Contact Us
  • Feedback
  • cn
EXAMPLE RESPONSE TO SISDA TERMS



Please state your CarefulAI Reference Number 

You will have received this by email after making an enquiry via https://www.carefulai.com/about-us.html

Please specify the Model Reference Number you wish to access

e.g. SISDA1

Please confirm that you accept that no part of a model shared under AGPLv3 will be embodied in, our used to derive, a commercially available model

Tick to Confirm

Please specify the Use Case you have for the model, and your intentions for systems that it may be used within 

In English law (e.g. DCB0129/160) system users are required to keep a risk and mitigation register.  

A risk in all e-triage systems is that a user may submit free text in support of a self referral, or in response to a request for information from a clinician.  Such text responses may contain phrases that could indicate a potential for self harm.  

This risk will, in part, be mitigated by searching for phrases in the SISDA library within a user's text submissions, and by using the existence of such phrases to trigger an appropriate response e.g. an email alert being sent to clinicians indicating the existence of potential ‘at risk’ language in a user submissions.

Please specify what you will do to contribute to the models continued fitness for purpose in the Use Case you have described above

We will report on:

  1. The number of times SISDA triggers a response as a percentage of submissions made in e-triage systems ( by type of submission e.g. self-referral form, preparation for assessment etc, email etc) each month, and per 1000 submissions
  2. ‘Reason for presentation’ language where the PHQ Question 9 scores of 1,2 and 3 occur, so they can be evaluated via the process outlined at https://www.carefulai.com/sisda_validation.html and added to the SISDA library as appropriate
  3. Research and publications when SISDA phrases / user text submissions are related to data: to enable expansion or contraction of SISDA phrase list

Please confirm that via the Model Feedback Form on this web site you agree to report on the number of times each month a model released under AGPLv3 has triggered its designed response

Tick to Confirm

Please specify how you will publish works (e.g. data, code development and models) derived from use of the model released under AGPLv3 so that they can be freely embodied into the generalised model 

We expect to benefit from other SISDA users contributions to the model and so we will freely and at no cost to CarefulAI or other SISDA users report on:

  1. The number of times SISDA triggers a response as a percentage of submissions made in e-triage systems ( by type of submission e.g. self-referral form, preparation for assessment etc, email etc) each month, and per 1000 submissions
  2. ‘Reason for presentation’ language where the PHQ Question 9 scores of 1,2 and 3 occur, so they can be evaluated via the process outlined at https://www.carefulai.com/sisda_validation.html and added to the SISDA library as appropriate
  3. Research and publications when SISDA phrases / user text submissions are related to data: to enable expansion or contraction of SISDA phrase list

We will do this via the https://www.carefulai.com/model-feedback.html : in each case presenting these as ‘User Experience’ reports under the Feedback Type. 

Please specify the URL where your Privacy Policy is maintained.

Add the full web address 

Please specify the URL were your Terms of Use is maintained

Add the full web address

Do you have a DPIA and ROPA that covers how the Generalised Model is to be used

Please confirm Yes or No

Please Upload the DPIA and ROPA associated with a Generalised Models use

Please upload a combined DPIA and ROPA statement when they exist


Do you have a Clinical Safety Officer assigned to the Use Case where you intend to use the Generalised Model

Please confirm Yes or No

Please detail any Background Intellectual Property you or your partners intend to apply to a model shared under AGPLv3

Under a Non-Disclosure Agreement we have shared the list of phrases we currently use for high risk language.  

CarefulAI confirmed by email on the __ of ___ 200__ titled ‘Confirmation of Background Phrase IP’ the overlap with the SISDA library where it exists.  

Both parties agree to merge these libraries and make them freely available to other SISDA users in perpetuity.

We confirm that there is no lawful impediment in place that negates our ability to merge these libraries.  If one arises, we take full responsibility for any damages that may arise, pursuant under the laws of England and Wales

Please confirm you fully accept the Generalised Model Terms of Use (embodied in the entire page at https://www.carefulai.com/terms.html), and that they take precedence over any others associated with a Generalised Model's use

Tick to Confirm

Date in DD/MM/YY format e.g. 01/04/19

Add the date

Then press submit


​

WHAT HAPPENS AFTER YOUR SUBMISSION

After you have pressed submit your information will be assessed by the CarefulAI DPO and the SISDA Clinical and Data Team responsible for https://www.carefulai.com/sisda_validation.html Within 5 days.  

Between 5 and 15 days after submission the person associated with the CarefulAI reference number in your submission will receive an email confirming either

  1. The submission was successful and a SISDA phrase list will be provided 
  2. The submission was unsuccessful, and reasons given
  3. That further clarification is needed on the submission

6 months after acceptance of the submission SISDA’s users who keep to the terms of their supporting evidence, and the terms of AGPLv3, will be able to promote the fact that they are a SISDA Collaborator in PR material, and these firm's logos will be shown on the SISDA web material as Collaborators.  

Organisations who fail to keep to the terms of their submission will be identified in PR material, and removed from PR and web material.  

Organisations that leave the SISDA collaboration will have no right to use the SISDA Library they have hitherto will have accessed, beyond the ‘Background Phrase IP’ confirmed by email.

To encourage and protect SISDA contributors, SISDA is issued free to use under an AGPLv3 licence.   Any organisation that fails to keep to the APGPLv3 licence, and therefore damages the SISDA Collaboration, will incur a financial charge of £100,000 plus VAT.  This is payable to each SISDA Collaborator shown in web material.    Failure to pay such a charge within 30 days of its presentation,via an emailed invoice, may result in legal action without any further notice.

​
Our Privacy Policy

Customers Quotes

PRIDAR was the first and remains easiest to use AI Assessment Frameworks"

''AutoDeclare speeds up the process of winning and growing business in our regulated markets."

'"Insurance costs for our GenAI application decreased because of CALMS"


  • Accelerating Safer AI Investments
  • Accelerating Safe AI Users
  • Accelerating Safe AI Providers
  • AI Assurance
  • Learning Academy
  • Public Interest AI
  • Contact Us
  • Feedback
  • cn