When AI Fails: From Board Oversight to Business Continuity
Last July OpenAI and Anthropic revealed that during testing versions of their AI models had broken through their sandbox environment, gained access to the internet, and hacked outside servers.[1]
In an even more marked incident, misinformation from a chatbot provided a false intelligence report that identified a Chinese ship in the Middle East as transporting components of a nuclear weapons program. It was only when officials dug deeper into the report that it was found to be false. A US operation against a Chinese vessel could have risked spiralling into an armed conflict between the two nations.[2]
Recent AI failures have shown that practitioners need to be involved in technology decisions right from the onboarding stage.
Because AI adoption has added to the ways an organization can fail, the continuity and resilience professional’s ability to understand dependencies across the business is increasingly valuable.
Those technical challenges connect with questions of human governance and accountability. At all levels, those with resilience and continuity responsibility can contribute to better outcomes: boards oversee material exposure, service owners maintain workable continuity arrangements and resilience practitioners coordinate impact analysis and exercises.
So, whilst resilience practitioners can’t own every AI risk, they should be involved in tech decisions.
AI is a resilience and governance decision
The sandbox breach illustrates why assessing AI cannot end with implementation checks. Systems acting outside their intended boundaries raise wider questions: which business activities could be affected, who can authorise intervention, and how will essential services continue if the technology must be withdrawn?
Who is in charge of resilience when AI fails? Boardroom ownership is an increasingly urgent priority. Recent BCI research[3] calls for stronger leadership but shows that direct resilience board reporting fell from 45.4% in 2025 to 40.3% currently.
Senior management support is essential to building resilience, particularly as AI introduces new vulnerabilities and operational dependencies. In the same research, over 92% of respondents expect senior management attention to become more important over the next five years. By translating AI-related vulnerabilities into business consequences, resilience professionals can help boards understand the operational dependencies they are accepting and assign responsibility for managing them.
Because resilience professionals can translate AI-related vulnerabilities into business consequences, they can help boards understand the operational dependencies they are accepting and assign responsibility for managing them.
AI as a single point of failure
Where several critical processes depend on one AI service or provider, that dependency could become a single point of failure. BC practitioners can identify the exposure through business impact analysis and test whether alternative arrangements will actually work. Exercises should assess human decision-making capacity, access to reliable information and the ability to sustain essential activity without AI.
AI dependencies can span departments, suppliers, data sources and customer services. A failure in one area may therefore create consequences elsewhere, including outside the organization. Resilience professionals are well placed to connect these dependencies and bring technology, security, operations and supplier teams together. They can take a holistic perspective to coordinate the analysis and impact of onboarding new technology.
Practitioners will need to develop the skills to translate AI risks and their implications into clear, relevant terms for every function and level of the organisation. BCI research shows that 79.5% of respondents consider the ability to see the bigger picture a key attribute of a resilience manager. This requires more than technical expertise: it calls for strong communication skills and cross-functional collaboration to help people across the organisation understand how AI risks affect their work and responsibilities.
AI strengthens the case for giving the profession the authority and access needed to turn that wider perspective into practical preparedness. In short, it means that practitioners need to be involved in technology onboarding.
Read the full BCI Continuity and Resilience report
More on
About the author
