Upskill / Business and Digital / Governing and certifying AI in aviation
Governing and certifying AI in aviation
Accountability, governance frameworks, safety assessment, data, certification and change control for AI functions. Part 3 of the AI in aviation series.
By Aerospace Garage Desk · Managing it · 50 minutes · 24 narrated pages
What you will be able to do
- ✓ Explain why accountability for AI functions stays with approval holders and named people, and sort the main AI instruments into binding rules, guidance and proposals as at October 2026.
- ✓ Design a company AI governance framework with a register, risk tiers, approval gates and independent sign-off, built into an existing aviation management system.
- ✓ Assess an AI function's safety using failure condition classes and an organisation's risk matrix, converting error rates into events per year and choosing barriers.
- ✓ Set data governance, records and portfolio monitoring for AI functions, and write an audit finding on an uncontrolled model update.
- ✓ Describe the certification and change-control route for airborne and ground AI functions, using the official findings on the Boeing 777 9M-MRG upset (ATSB 200503722).
- ✓ Lead AI adoption with an honest business case, competence for approvers, early engagement with the authority and a concise board report.
Course content
Session 1 · Who governs AI in an aviation organisation
1.1
- The question every approval holder must answer · free preview
- What binds you today, and what is still a proposal
- Standards: the toolkit for your evidence
- Worked example: drawing Port Hills Aero's accountability map
Session 2 · Building a company AI governance framework
1.2
- Eight elements, built into what you already have
- The register and the risk tiers
- Approval gates and who signs
- Worked example: hunting the outliers in Port Hills Aero's register
Session 3 · Safety assessment of AI functions
1.3
- The classic yardstick: severity against probability
- Where machine learning strains the method
- Ground tools: assessing them in the safety management system
- Worked example: pricing the misses of a borescope assistant
Session 4 · Data governance and records
1.4
- Who owns the data that writes the model
- Records: proving what was decided and why
- Monitoring a portfolio in service
- Worked example: writing an audit finding on an unreviewed model update
Session 5 · Certification and change control
1.5
- Certifying an airborne AI function
- Changing a model: classify, test, approve
- When things go wrong in service
- From the investigation file: the latent software anomaly behind 9M-MRG's upset
Session 6 · Leading AI adoption
1.6
- The business case, including the cost of assurance
- Competence, culture and the authority
- Reporting to the board
- Worked example: taking the borescope assistant through approval
Who this course is for
- Accountable managers and compliance heads setting up an AI register and approval gates
- Safety managers assessing AI functions with failure condition classes and risk matrices
- Auditors who must write a finding on an uncontrolled model update
You need: Using AI tools safely at work in aviation (Part 2 of this series) and experience of a management system.
Upskill courses are for professional development and are not approved by any aviation authority. Learner Terms
Course content © Aerospace Garage Limited. All rights reserved.