Source · Select Committees · Science and Technology Committee (Commons)
Third Report - Governance of artificial intelligence (AI)
Science and Technology Committee (Commons)
HC 38
Published 28 May 2024
Source-linked government response
1st Special Report – Governance of artificial intelligence (AI): Government Response · published 10 Jan 2025
How to read the colours. These are AI-assisted assessments of the position taken in the published government response. They are not official government labels and do not show whether promised action was delivered.
Definitions for the six response statuses
- Accepted
- Accepts the item, commits to concrete action, or says existing work already addresses it.
- Accepted in Part
- Accepts only part of the item or offers narrower action.
- Acknowledged
- Recognises the issue without making a specific commitment.
- Deferred
- Leaves action to a future process or redirects responsibility elsewhere.
- Not Addressed
- Does not engage with the substance of the item.
- Rejected
- Explicitly declines or disagrees with the item.
- Response status not verified
- Our records do not establish the response status. A missing link or passed expected date does not show that no response was published.
- Response attribution not verified
- A response has not been verified against this item. This is a limitation of our records, not a finding about the government.
Read the published response text for the full context.
Recommendations & Conclusions
32
Recommendation
Not Addressed
Require robust, independent testing and performance analysis for AI models prior to deployment.
Recommendation
AI can entrench and accelerate existing biases. The current Government, future administrations and sectoral regulators should require deployers of AI models and tools to submit them to robust, independent testing and performance analysis prior to deployment. (Paragraph 140) 56 Governance of artificial intelligence (AI)
Read the full government response
56 Governance of artificial intelligence (AI)
Government response · AI-generated summary
The government's response is boilerplate text and does not address the recommendation to require robust, independent testing of AI models for bias prior to deployment.
Department for Science, Innovation and Technology
View Details →