GAICC AI Conference & Awards 2026 "Governing the Future – Building Responsible, Safe and Human-centric AI"
GAICC AI Conference & Awards 2026 "Governing the Future – Building Responsible, Safe and Human-centric AI"
Conventional AI governance asks whether a system’s output is accurate, fair and lawful. Agentic governance asks what a system may do and on whose authority. Robotic AI governance asks the hardest question of the three:
What is this system permitted to do with mass, force and speed, among people who did not choose to be near it — and what evidence proves it will stop when it must?
RAIGP is the first GAICC credential written for systems whose failures are measured in kinetic energy.
Students
The GAICC Robotic AI Governance Professional (RAIGP) is an advanced practitioner credential for professionals who authorise, oversee, assure and stop AI systems that move through the physical world.
That scope is deliberately broad, because the governance problem is the same across all of it: industrial and collaborative robots, mobile robots and fleets, service and humanoid robots, medical and surgical robots, agricultural machines, autonomous vehicles and uncrewed aircraft.
A model that produces a wrong answer creates an informational harm you can usually correct. A robot that produces a wrong action creates a physical harm you cannot. Reversibility, detectability and time-to-harm become the primary governance variables, and kinetic energy, reach, payload and speed become governance inputs rather than engineering footnotes.
RAIGP is built around the decisions that follow from that: accept or refuse residual risk, authorise or withhold deployment, stop or continue, notify or hold, promote or demote autonomy, sign or decline the declaration of conformity.
RAIGP sits at the junction of two established fields that have not yet been joined:
Learned control policies, robot foundation models, fleet learning and teleoperation break both assumptions at once.
RAIGP holders are the professionals who can read a functional safety claim and an AI impact assessment in the same meeting, and know which one is missing.
The credential extends the GAICC AI Governance Framework, Maturity Model and Implementation Roadmap into embodied operation. It assumes the fundamentals taught in CPAIG and the authority and control concepts taught in AAIGP, and spends all 32 hours on the physical safety, assurance, security, liability and regulatory problems created by autonomy with a body.
Frameworks, standards and regulation applied throughout the programme: EU AI Act (as amended by Regulation (EU) 2026/1744) · EU Machinery Regulation 2023/1230 · Product Liability Directive (EU) 2024/2853 · Cyber Resilience Act · ISO 10218:2025 · ISO 13482 · ISO 3691-4 · ISO 12100 · ISO 13849-1 · IEC 61508 · UL 4600 · ISO 21448 · ISO/IEC 42001 · ISO/IEC 42005 · ISO/IEC 23894 · ISO/IEC TR 5469 · ISO/IEC TS 8200 · IEC 62443 · NIST AI RMF · the GAICC AI Governance Framework
Issued By Global AI Certification Council
Functional safety standards let you claim a performance level or safety integrity level for a safety function because that function behaves the same way every time. Insert a learned component and the claim stops meaning what it used to mean — and none of ISO 13849-1, IEC 62061 or IEC 61508 currently addresses a safety function whose behaviour evolves.
| When | What applies |
|---|---|
| 11 September 2026 | EU Cyber Resilience Act vulnerability and incident reporting applies to connected robots |
| 9 December 2026 | EU Product Liability Directive 2024/2853 applies — software and AI treated as products, with presumptions of defect and substantial modification through updates or learning |
| 20 January 2027 | EU Machinery Regulation 2023/1230 applies — machinery with self-evolving safety components attracts notified body assessment |
| 11 December 2027 | Cyber Resilience Act applies in full — security by design, conformity assessment, software bill of materials |
| 2 December 2027 | EU AI Act Annex III high-risk obligations apply |
| 2 August 2028 | EU AI Act Annex I obligations apply to AI embedded in regulated products; Machinery Regulation AI requirements via delegated act |
| Dimension | GAICC RAIGP | GAICC AAIGP | GAICC CPAIG | Functional safety credentials | General AI governance credentials |
|---|---|---|---|---|---|
| Governing question | What may this system do with mass, force and speed, among people — and what proves it will stop? | What may this system do, on whose authority, and how is it stopped? | Is this AI system lawful, fair and well governed? | Does this machine meet the required performance level or safety integrity level? | Is this AI system lawful and well documented? |
| Physical safety and the machinery regime | Full domain — ISO 12100, ISO 10218:2025, ISO 13482, ISO 3691-4, functional safety claims, Machinery Regulation, workplace safety law | Not addressed | Not addressed | Core content, engineering depth, no AI or learning | Not addressed |
| Learned control and robot foundation models | Full domain — learned policies, vision-language-action models, ISO/IEC TR 5469 configurations, deployer fine-tuning, fleet learning | Agent architectures and orchestration for software agents | Introductory coverage within technical foundations | Rarely addressed; assumes deterministic control | Rarely addressed |
| Human oversight | Full domain — stop and fallback, minimal risk condition, teleoperation, collaborative operation, human factors, public and care settings | Approval gates, circuit breakers, kill switches for software agents | Human oversight requirements and override principles | Emergency stop and safeguarding design | "Human in the loop" stated as a principle |
| Assurance and safety case | Full domain — UL 4600, GSN, AMLAS and SACE, scenario-based testing, simulation credibility, independent assessment | Trajectory evaluation, agentic red teaming, assurance cases | Commissioning and interpreting model safety testing | Verification against safety standards; safety case in UL 4600 programmes only | Limited |
| Cybersecurity, data and supply chain | Full domain — IEC 62443, Cyber Resilience Act, robot attack surfaces, sensor and teleoperation data, provenance and import controls | Non-human identity, tool grants, exfiltration through action | Data protection and third-party governance at principle level | Cybersecurity as a machinery requirement, limited depth | Privacy focus |
| Incident, liability and insurance | Full domain — physical incident response, recalls, workplace and product regulators, Product Liability Directive 2024/2853, insurance prerequisites | Agentic incident response, notification, supply chain liability | Incident and notification obligations | Accident investigation at engineering level | Documentation obligations |
| Regulation | AI Act applied through the machinery, medical, vehicle and aviation regimes; sector regimes; US, China, Korea, Japan and other jurisdictions | AI Act, GPAI and automated decision-making rights applied to agents | Global AI law landscape | Machinery and product safety law | AI law landscape |
| Training requirement | 32 hours | 32 hours | 28–30 hours | Typically 4–5 days | Varies |
| Examination | 100 scenario items · 2.5 hours · 70% pass | 100 scenario items · 2.5 hours · 70% pass | 80 scenario items · 2 hours · 70% pass | Typically 2–6 hours, engineering calculation, 80% pass in some schemes | Typically 100 items · 3 hours · scaled pass |
You do not need to be an engineer. RAIGP asks you to read, challenge and act on engineering evidence — risk assessments, functional safety claims, safety cases — not to produce it or perform calculations.
ISO 8373 robot classification · Robot and fleet registry design · Safety layer independence assessment · Operational design domain specification · ISO 12100 risk assessment · Residual risk acceptance · ISO 10218:2025 and ISO 3691-4 application · Performance level and SIL claim review · IEC 60204-1 stop categories · Machinery Regulation conformity assessment · CE marking and notified body decisions · Lockout-tagout and ISO 45001 integration
Embodied AI risk taxonomy · Kinetic energy and reversibility classification · STPA and unsafe control actions · HAZOP, FMEA, FTA and HARA · Triggering-condition inventories · Distribution shift and edge-case assessment · Prompt injection into language-directed robots · Embodied AI impact assessment · Minimal risk condition and manoeuvre design · Geofencing and safety envelopes · Teleoperation operating rules · Collaborative operation method selection · Anthropomorphism and dignity governance
UL 4600 safety cases · GSN and CAE structured argument · AMLAS and SACE for machine-learnt components · Scenario-based testing and simulation credibility · Semantic safety and refusal evaluation · Physical AI red teaming · Over-the-air update change control · Independent safety case assessment · IEC 62443 zones and conduits · Robot attack-surface threat modelling · Cyber Resilience Act obligations · Robot data lifecycle and DPIA · Log integrity and evidence preservation · Software bill of materials and provenance
Fleet operating models and safety performance indicators · Physical incident response sequencing · Software recalls and field corrective actions · Multi-regulator notification · Product Liability Directive 2024/2853 application · Contractual allocation and indemnities · Insurance cover and underwriting prerequisites · STPA-based post-incident review · AI Act Annex I and Annex III classification for robots · Sector regimes (medical, vehicle, aviation, agriculture) · Multi-jurisdiction mapping · ISO/IEC 42001 integration with ISO 45001, ISO 9001, ISO/IEC 27001 and IEC 62443 · Board reporting on physical autonomy
| Module | Focus | Applied activity |
|---|---|---|
| M00 Orientation | GAICC, the credential, the examination and the embodied shift | Scene setter: three real robot failures and which domain would have caught each |
| M01 Classifying robots | ISO 8373 vocabulary, autonomy and environment, registry entries | Classification clinic: tier ten real deployments by class, environment and regime |
| M02 The robot stack and the safety layer | Where AI decides vs where deterministic control decides; kinetic energy as a governance variable | Architecture teardown: is the safety controller actually independent of the AI? |
| M03 Learned policies and robot foundation models | Vision-language-action models, ISO/IEC TR 5469 configurations, deployer fine-tuning, ODD | Write the operational design domain for a humanoid moving totes in a shared aisle |
| M04 Fleets, connectivity and teleoperation | Standalone, fleet-managed, cloud-dependent and edge-only architectures | Accountability map: who is responsible when the link drops? |
| M05 Machinery risk assessment | ISO 12100, ISO 10218-1/-2:2025, ISO 3691-4, R15.08, ISO 13482 | Challenge a real collaborative robot risk assessment, then sign or refuse in writing |
| M06 Reading functional safety claims | Performance levels, SILs, IEC 60204-1 stop categories | Reading exercise: find the unevidenced claim in a supplier safety data sheet |
| M07 Conformity assessment and market placement | Machinery Regulation 2023/1230, manufacturer vs integrator vs deployer, CE marking | Decision tree: who signs the declaration of conformity after the deployer retrains it? |
| M08 Workplace health and safety | Lockout-tagout, ISO 45001, maintenance and teach modes, worker consultation | Case study: write the controls that would have prevented a maintenance-mode fatality |
| Module | Focus | Applied activity |
|---|---|---|
| M09 The embodied AI risk taxonomy | Perception, planning, control, oversight, cyber and organisational failure | Sort 20 real incidents onto the taxonomy, then to the owning control |
| M10 Hazard analysis for autonomy | STPA, HAZOP, SOTIF, triggering-condition inventories | STPA workshop: a mobile robot sharing an aisle with pickers |
| M11 Learned-policy and foundation-model risk | Distribution shift, adversarial input, risk appetite for physical autonomy | Red-flag drill: find the injection paths into a language-directed service robot |
| M12 Embodied AI impact assessment | ISO/IEC 42005 extended to physical harm; bystanders and vulnerable people | Template build: impact assessment for a patrol robot in a public shopping centre |
| M13 Robotic risk in enterprise risk management | Fleet risk register, severity classification, escalation triggers | Worked example: a fleet risk register row end to end, then board aggregation |
| M14 Meaningful human oversight | In / on / out of the loop, Article 14, ISO/IEC TS 8200 controllability | Critique four oversight designs: which survives a regulator, which survives a coroner |
| M15 Stop, fallback and containment | Emergency and protective stop, minimal risk condition, geofencing | Tabletop: stop a humanoid mid-task near people; define a robotaxi's minimal risk condition |
| M16 Teleoperation and remote assistance | Ratios, latency limits, handover, operator fatigue, honest disclosure | Write the teleoperation operating rules for a home humanoid service |
| M17 Human-robot interaction and public settings | Collaborative methods, work pace, anthropomorphism, dignity | Build an autonomy promotion ladder from segregated to public operation |
| Module | Focus | Applied activity |
|---|---|---|
| M18 Building and challenging the safety case | UL 4600, GSN, CAE, AMLAS, SACE | Safety case canvas, peer reviewed — each reviewer must find one unevidenced claim |
| M19 Verification, validation and simulation credibility | Acceptance criteria for physical behaviour, digital twins, the sim-to-real gap | Write a V&V specification a vendor can be held to |
| M20 Evaluation and red teaming of physical AI | Semantic safety evaluation, sensor spoofing, rules of engagement | Red team planning, then read a findings report and decide go or no-go |
| M21 Change control, fleet learning and independent assessment | OTA updates, independent verification before closure, certification types | Change-control drill: a policy update changes school-zone behaviour — what must be verified, by whom? |
| M22 Robot cybersecurity, physical security and misuse | IEC 62443, ISO 10218:2025 cyber requirement, Cyber Resilience Act, tampering | Threat model a quadruped and a humanoid against IEC 62443 zones |
| M23 The robot data lifecycle and evidence integrity | Sensor and biometric data, DPIA, minimum event logging, retention | Write a logging specification, then test it against three audit questions and a disclosure request |
| M24 Supply chain, provenance and third parties | Due diligence, software bills of materials, import/export controls | Due diligence questionnaire and contract markup for a humanoid supplier |
| M24R Day 3 consolidation | Domains V and VI reviewed against the examination blueprint | Timed domain review with rationale walkthrough |
| Module | Focus | Applied activity |
|---|---|---|
| M25 Fleet operations and physical incident response | Scene safety, isolation, evidence preservation, fleet-wide containment | Live tabletop: a robot has struck a worker; the same release runs on 400 robots across six sites |
| M26 Notification, recalls, liability and insurance | Multi-regulator reporting, software recalls, Product Liability Directive, underwriting | Contract and insurance clinic: allocate liability for a learned-policy update that caused harm |
| M27 Learning from incidents | Control-structure-focused post-incident review, systemic findings | STPA-based review of a fix declared closed that then failed again |
| M28 AI regulation, sector regimes and jurisdictions | AI Act Annex I and III, medical, vehicle, aviation, agricultural regimes; US, China, Korea, Japan and more | Jurisdiction mapping for one humanoid platform across a factory, a hospital and a shopping centre in three regions |
| M29 Management system mapping and board advisory | ISO/IEC 42001 mapped to the robot lifecycle; Maturity Model; portfolio scaling | Board pack build: five slides on the robot portfolio, presented and challenged |
| MOCK Full mock examination | 100-item timed mock under examination conditions | Rationale-led debrief and domain-level performance report |
Total Domains
8 Domains · 100% Weighting
RAIGP certifies the ability to govern learned behaviour and machinery safety together. Neither an AI governance credential nor a functional safety credential currently assesses that combination.
Built to the international standard for bodies certifying persons, with certification decisions taken independently of training delivery and independently of the scoring platform.
Every item puts you in a decision seat. No question tests definition recall alone, no question requires engineering calculation, and no question can be answered from vendor terminology.
All 40 competencies are mapped to the modules that teach them. What you study is exactly what is assessed — and the blueprint is public before you enrol.
A digital badge and verifiable certificate for LinkedIn, CVs and professional profiles, plus listing in the GAICC global credentials register.
RAIGP is designed to sit alongside CPAIG and AAIGP, and alongside functional safety credentials such as CMSE, TÜV FS Engineer, CFSE and UL-CASP. It does not replace engineering competence; it governs the system around it.
There is one hard requirement: 32 hours of structured RAIGP training. Professional experience is advisory and is not a barrier to access, and no engineering qualification is needed.
Choose any one of the three recognised routes — GAICC self-paced, a GAICC-authorised instructor or training partner, or a GAICC-recognised institution. All three are equivalent for eligibility.
Read the Examination Content Outline in full, work the module quizzes and the full mock under timed conditions, and use the domain-level report to target revision.
Book through hub.gaicc.org and run the system check well ahead of the day. 100 scenario-based items in 150 minutes, closed book, from a quiet private space anywhere in the world. Your provisional result and domain feedback appear on screen at completion.
The GAICC Certification Decision Authority reviews the provisional result and confirms the decision within 72 hours. Successful candidates receive a verifiable certificate and a digital badge.
| Total questions | 100 scenario-based items (88 scored + 12 unscored pilot) |
| Format | Scenario-based multiple choice, four options (A–D), single best answer |
| Duration | 150 minutes (2 hours 30 minutes), running continuously with no scheduled break |
| Pass mark | 70% of scored items — a minimum of 62 of 88 |
| Scoring | Scaled 100–500; 350 is the passing threshold, confirmed through modified-Angoff standard setting |
| Negative marking | None — answer every question |
| Cognitive split | Understanding 15% · Application 45% · Analysis and Evaluation 40% |
| Reference material | Closed book. No external reference material permitted. No item requires calculation |
| Delivery | AI-proctored online on the GAICC hub (hub.gaicc.org), global browser access |
| Language | English (other languages on request) |
| Results | Provisional on screen at completion, with domain-level performance feedback |
| Certification decision | Confirmed by the Certification Decision Authority within 72 hours |
RAIGP is an advanced credential, but access is deliberately open. There is one hard requirement: completion of 32 hours of structured RAIGP training by any of the three recognised routes, before you sit the examination.
The certification is open to practitioners in AI governance, robotics and autonomous systems, machinery and functional safety, health and safety, risk, compliance, audit, security, product compliance, insurance or law.
| Route | Description |
|---|---|
| 1. GAICC self-paced course | The 32-hour RAIGP course completed in your own time on the GAICC hub |
| 2. GAICC-authorised Certified Instructor or Authorised Training Partner | Instructor-led delivery of the 32-hour programme by an authorised partner or an authorised GAICC Certified Instructor |
| 3. A GAICC-recognised institution | 32 hours of RAIGP training delivered by an institution recognised by GAICC for this purpose |
Strongly recommended, not required: the GAICC Certified Professional in AI Governance (CPAIG) or the GAICC Agentic AI Governance Professional (AAIGP), or an equivalent AI governance credential. RAIGP assumes working knowledge of AI governance fundamentals and does not re-teach them.
Helpful: professional experience in robotics or autonomous systems delivery, machinery or functional safety, health and safety, AI governance, risk, compliance, audit, security, product compliance, insurance or law. Experience is advisory. It supports readiness and is not a barrier to access.
Not required: an engineering qualification. The examination asks you to read and challenge engineering evidence, not to produce it — and no item requires calculation.
A note on your starting point. Candidates holding a machinery or functional safety credential (CMSE, TÜV Functional Safety Engineer, CFSE, CFSP, UL-CASP) will find Domain II familiar and should expect Domains I, III, V and VIII to carry the AI governance content they have not met. Candidates from an AI governance background should expect the reverse, and give extra time to Domains II, IV, VI and VII.
⭐⭐⭐⭐⭐
⭐⭐⭐⭐⭐
⭐⭐⭐⭐⭐
⭐⭐⭐⭐⭐
⭐⭐⭐⭐⭐
⭐⭐⭐⭐⭐
RAIGP is 32 contact hours carrying 32 CPD, structured as four eight-hour days. Each day covers two examination domains and closes with a consolidation review against the blueprint. Day 4 ends with a full-length timed mock examination and a rationale-led debrief.
You can complete the hours instructor-led, or self-paced on the GAICC hub — whichever suits your role and time zone.
US$898
US$1075
US$99 is included in the above member price.
| Competency | Hours per three-year cycle |
|---|---|
| Technology | 27 hours |
| Strategy | 21 hours |
| Humanity | 12 hours |
| 60 hours |
Hours may be earned through robot and autonomous system authorisation, risk and hazard assessment, oversight and safety case work, evaluation, audit, incident response, and standards or policy contribution — as well as recognised learning.
Renewal fee: US$69 for each three-year cycle. A three-month grace period applies after expiry.
Where CPD evidence is insufficient, where the certification has lapsed beyond the grace period, or where the scheme changes materially, re-examination is required.
Use of the certificate and GAICC marks: certified persons may state that they hold the certification while it is valid and use the GAICC certification mark in accordance with the Use of Certificates and Marks procedure. The certification applies to the named individual, is not transferable, and must not be used misleadingly, to imply accreditation that has not been granted, or to imply engineering or functional safety competence that the credential does not certify. Lapsed or withdrawn certificants must cease use of the certification and the marks.
Hours may be earned through robot and autonomous system authorisation, risk and hazard assessment, oversight and safety case work, evaluation, audit, incident response, and standards or policy contribution — as well as recognised learning.
Renewal fee: US$69 for each three-year cycle. A three-month grace period applies after expiry.
Where CPD evidence is insufficient, where the certification has lapsed beyond the grace period, or where the scheme changes materially, re-examination is required.
| Principle | Obligation |
|---|---|
| 1. Integrity and Fairness | Act honestly in all professional activities without bias, conflict of interest or misrepresentation of competence. Disclose potential conflicts proactively. |
| 2. Proportionate Authority | Never recommend or approve a level of agent autonomy that exceeds the assurance evidence held. Where authority and evidence diverge, say so in writing. |
| 3. Human Rights and Wellbeing | Ensure autonomous systems under your governance influence respect human dignity, fairness, privacy and non-discrimination — including for people who never interact with the system. |
| 4. Transparency and Accountability | Promote explainable and reconstructable agent behaviour. Ensure a named human remains accountable for every autonomous action. |
| 5. Data Protection and Privacy | Uphold confidentiality and data-protection principles across agent memory, tool access and trace data, consistent with applicable laws and international standards. |
| 6. Professional Competence | Maintain current knowledge in a field that is changing faster than its regulation, through continuous learning and active engagement with evolving standards. |
| 7. Reporting and Mitigation of Misuse | Take appropriate action when encountering unsafe autonomy, suppressed incidents, disabled controls or violations of applicable regulation or professional standards. |
| 8. Global Responsibility | Recognise that autonomous systems act across borders. Apply governance standards that protect individuals regardless of jurisdiction. |
Director at the Global AI Certification Council (GAICC) and PM Training School
A globally certified instructor in ISO/IEC, PMI®, TOGAF®, SAFe®, and Scrum.org disciplines. With over three years’ hands-on experience in ISO/IEC 42001 AI governance, he delivers training and consulting across New Zealand, Australia, Malaysia, the Philippines, and the UAE, combining high-end credentials with practical, real-world expertise and global reach.
Self-Paced Course (Certification Exam included)
US$898
US$1075
GAICC Annual Membership US$99 is included in the above member price.
