Synoptix’ Callum Cockburn has contributed to a new Tony Blair Institute report examining how effective assurance can support safer and accelerated AI adoption.
A new report from the Tony Blair Institute for Global Change explores how stronger AI-assurance ecosystems can help organisations adopt artificial intelligence safely and with greater confidence.
AI-Assurance Ecosystems for Safer and Accelerated AI Adoption considers the technical, regulatory and institutional capabilities needed to assess AI systems, manage risk and support responsible innovation.
Synoptix’ Callum Cockburn, Senior Technical Innovation Manager, is recognised as a contributor to the report. His input covered several principles central to our approach, including lifecycle and system-level assurance, assured development, outcome-based regulation and structured assurance cases.
Together, these principles support an important message: assurance should help organisations design and improve better AI-enabled systems, rather than simply demonstrate compliance once they have been built.
Assurance throughout the lifecycle
AI assurance is sometimes treated as an activity that takes place towards the end of development. However, decisions made much earlier can determine whether a system can ultimately be assured.
As the report explains:
“Critically these activities should operate throughout the AI lifecycle, with assurance evidence informing system design and development, as well as decisions about procurement, deployment and ongoing operation, rather than functioning primarily as a retrospective compliance exercise.”
This places assurance within the development process itself, where evidence can guide decisions, identify weaknesses and improve systems as they evolve.
Effective assurance must also look beyond the AI model. AI-enabled capabilities depend on interactions between models, data, software, safeguards, human operators and the operating environment. System-level assurance considers how these elements work together under real conditions, particularly in high-stakes and safety-critical contexts.
Focusing on outcomes and evidence
The report also considers outcome-based regulation, which defines the safety, reliability or performance a system must achieve without prescribing a single method for demonstrating it.
The UK Military Aviation Authority is highlighted as an example. Its approach requires AI-enabled air systems to be assured to a level of confidence proportionate to the risk of failure, with performance expectations defined for the relevant operating context.
The report describes this as an emphasis on “demonstrated outcomes over prescribed process”. Allowing the same outcome to be evidenced in different ways could make assurance evidence more portable between regulatory environments and support greater cross-border acceptance.
For high-stakes AI systems, assurance cases can also provide a structured way to connect claims about a system with the reasoning and evidence supporting them.
As the report states:
“Independent third-party assurance providers then play a central role in assessing the claims and evidence. Such an approach preserves the stability of regulation while allowing assurance practice to move with the technology.”
Commenting on the report, Callum said:
“It was a pleasure to contribute to Elizabeth and Darcy’s report, and to discuss these issues directly with Elizabeth. Its central message is an important one: assurance should not be something we apply once an AI system has already been built. Used throughout development and at system level, it helps organisations address risk early, improve their designs and adopt AI with justified confidence.”
The report makes an important contribution to the discussion around trusted AI adoption. By building assurance into the lifecycle, evaluating AI at system level and focusing on meaningful outcomes, organisations can create stronger foundations for deploying AI safely and with confidence.
Read the full Tony Blair Institute report: here
Find out more about Synoptix’ approach to AI and system assurance: here
Building assurance into AI from the outset
Synoptix’ Callum Cockburn has contributed to a new Tony Blair Institute report examining how effective assurance can support safer and accelerated AI adoption.
A new report from the Tony Blair Institute for Global Change explores how stronger AI-assurance ecosystems can help organisations adopt artificial intelligence safely and with greater confidence.
AI-Assurance Ecosystems for Safer and Accelerated AI Adoption considers the technical, regulatory and institutional capabilities needed to assess AI systems, manage risk and support responsible innovation.
Synoptix’ Callum Cockburn, Senior Technical Innovation Manager, is recognised as a contributor to the report. His input covered several principles central to our approach, including lifecycle and system-level assurance, assured development, outcome-based regulation and structured assurance cases.
Together, these principles support an important message: assurance should help organisations design and improve better AI-enabled systems, rather than simply demonstrate compliance once they have been built.
Assurance throughout the lifecycle
AI assurance is sometimes treated as an activity that takes place towards the end of development. However, decisions made much earlier can determine whether a system can ultimately be assured.
As the report explains:
“Critically these activities should operate throughout the AI lifecycle, with assurance evidence informing system design and development, as well as decisions about procurement, deployment and ongoing operation, rather than functioning primarily as a retrospective compliance exercise.”
This places assurance within the development process itself, where evidence can guide decisions, identify weaknesses and improve systems as they evolve.
Effective assurance must also look beyond the AI model. AI-enabled capabilities depend on interactions between models, data, software, safeguards, human operators and the operating environment. System-level assurance considers how these elements work together under real conditions, particularly in high-stakes and safety-critical contexts.
Focusing on outcomes and evidence
The report also considers outcome-based regulation, which defines the safety, reliability or performance a system must achieve without prescribing a single method for demonstrating it.
The UK Military Aviation Authority is highlighted as an example. Its approach requires AI-enabled air systems to be assured to a level of confidence proportionate to the risk of failure, with performance expectations defined for the relevant operating context.
The report describes this as an emphasis on “demonstrated outcomes over prescribed process”. Allowing the same outcome to be evidenced in different ways could make assurance evidence more portable between regulatory environments and support greater cross-border acceptance.
For high-stakes AI systems, assurance cases can also provide a structured way to connect claims about a system with the reasoning and evidence supporting them.
As the report states:
“Independent third-party assurance providers then play a central role in assessing the claims and evidence. Such an approach preserves the stability of regulation while allowing assurance practice to move with the technology.”
Commenting on the report, Callum said:
“It was a pleasure to contribute to Elizabeth and Darcy’s report, and to discuss these issues directly with Elizabeth. Its central message is an important one: assurance should not be something we apply once an AI system has already been built. Used throughout development and at system level, it helps organisations address risk early, improve their designs and adopt AI with justified confidence.”
The report makes an important contribution to the discussion around trusted AI adoption. By building assurance into the lifecycle, evaluating AI at system level and focusing on meaningful outcomes, organisations can create stronger foundations for deploying AI safely and with confidence.
Read the full Tony Blair Institute report: here
Find out more about Synoptix’ approach to AI and system assurance: here
.png)
Building assurance into AI from the outset
Synoptix’ Callum Cockburn has contributed to a new Tony Blair Institute report examining how effective assurance can support safer and accelerated AI adoption.
A new report from the Tony Blair Institute for Global Change explores how stronger AI-assurance ecosystems can help organisations adopt artificial intelligence safely and with greater confidence.
AI-Assurance Ecosystems for Safer and Accelerated AI Adoption considers the technical, regulatory and institutional capabilities needed to assess AI systems, manage risk and support responsible innovation.
Synoptix’ Callum Cockburn, Senior Technical Innovation Manager, is recognised as a contributor to the report. His input covered several principles central to our approach, including lifecycle and system-level assurance, assured development, outcome-based regulation and structured assurance cases.
Together, these principles support an important message: assurance should help organisations design and improve better AI-enabled systems, rather than simply demonstrate compliance once they have been built.
Assurance throughout the lifecycle
AI assurance is sometimes treated as an activity that takes place towards the end of development. However, decisions made much earlier can determine whether a system can ultimately be assured.
As the report explains:
“Critically these activities should operate throughout the AI lifecycle, with assurance evidence informing system design and development, as well as decisions about procurement, deployment and ongoing operation, rather than functioning primarily as a retrospective compliance exercise.”
This places assurance within the development process itself, where evidence can guide decisions, identify weaknesses and improve systems as they evolve.
Effective assurance must also look beyond the AI model. AI-enabled capabilities depend on interactions between models, data, software, safeguards, human operators and the operating environment. System-level assurance considers how these elements work together under real conditions, particularly in high-stakes and safety-critical contexts.
Focusing on outcomes and evidence
The report also considers outcome-based regulation, which defines the safety, reliability or performance a system must achieve without prescribing a single method for demonstrating it.
The UK Military Aviation Authority is highlighted as an example. Its approach requires AI-enabled air systems to be assured to a level of confidence proportionate to the risk of failure, with performance expectations defined for the relevant operating context.
The report describes this as an emphasis on “demonstrated outcomes over prescribed process”. Allowing the same outcome to be evidenced in different ways could make assurance evidence more portable between regulatory environments and support greater cross-border acceptance.
For high-stakes AI systems, assurance cases can also provide a structured way to connect claims about a system with the reasoning and evidence supporting them.
As the report states:
“Independent third-party assurance providers then play a central role in assessing the claims and evidence. Such an approach preserves the stability of regulation while allowing assurance practice to move with the technology.”
Commenting on the report, Callum said:
“It was a pleasure to contribute to Elizabeth and Darcy’s report, and to discuss these issues directly with Elizabeth. Its central message is an important one: assurance should not be something we apply once an AI system has already been built. Used throughout development and at system level, it helps organisations address risk early, improve their designs and adopt AI with justified confidence.”
The report makes an important contribution to the discussion around trusted AI adoption. By building assurance into the lifecycle, evaluating AI at system level and focusing on meaningful outcomes, organisations can create stronger foundations for deploying AI safely and with confidence.
Read the full Tony Blair Institute report: here
Find out more about Synoptix’ approach to AI and system assurance: here
.png)
