# What if your AI model goes dark? Business continuity for AI dependencies

Source: https://perfraction.com/blog/ai-model-outage-business-continuity/
Author: Joshua Woo, Perfraction
Published: 2026-09-16
Topics: AI governance, Business continuity, Third-party risk

## Short answer

Treat the AI model as a critical third-party dependency that can become unavailable or less capable without notice, for reasons outside your contract. Map which workflows stop if it goes dark, name a fallback model or manual process for each, test the switch, and cover availability in your outsourcing and third-party risk framework.

## Key takeaways

- Model risk is not only accuracy risk. A model can also be unavailable.
- In June 2026, Anthropic suspended access to its two most capable models for nineteen days to comply with US export controls.
- Capability can shrink without disappearing, for example when a provider reduces a context window.
- Export controls and government orders are now operational inputs for AI users.
- Most firms have a BCP for the data centre. Very few have one for the model.

## Has a frontier AI model really gone dark?

Yes. On 12 June 2026, Anthropic suspended access to its two most capable models to comply with US export controls. Access came back on 1 July. Nineteen days.

Nothing was wrong with the model. Nobody breached a contract. No invoice went unpaid. The capability stopped being available, for reasons that had nothing to do with the customer, the vendor relationship or anyone's risk register.

## What should this change in AI governance?

1. **Model risk isn't only accuracy risk.** Most AI governance frameworks treat the model as something that might be wrong. It is also something that might not be there. Those need different controls.
2. **Capability can shrink without disappearing.** In June 2026 a Singapore lab cut a model's context window from 1 million tokens to 256,000. Nothing was withdrawn. Workflows built on the larger number quietly stopped fitting.
3. **Your dependency is geopolitical.** Export controls, national security reviews and compute rationing now feed into your operations, whether or not they appear in your third-party risk assessment.

## How do you plan for an AI model outage?

1. **Map the dependency.** List every workflow that calls an external model, and mark which ones are business-critical.
2. **Name the fallback.** For each critical workflow, name an alternative model or a manual process.
3. **Build the switch before you need it.** An abstraction layer lets you change providers by configuration, not by rebuilding.
4. **Test it.** Run a continuity exercise the way you would for a data centre failure. Time the switch and check output quality.
5. **Write it into contracts and frameworks.** Cover availability, notice and suspension in vendor terms, and add model unavailability to your outsourcing and third-party risk framework.

## What is the honest test?

If the model you rely on went dark tomorrow morning for nineteen days, what stops? And who in your organisation already knows the answer?

## Frequently asked questions

### Does an AI model outage count as a third-party risk event?

Yes. If a business process depends on an external model, loss of that model is a third-party service disruption. Financial institutions should check how their outsourcing and third-party risk framework treats a critical provider becoming unavailable by government order.

### How do you build a fallback for an AI model?

Route requests through an abstraction layer that can switch providers, keep a tested alternative model for critical workflows, and document a manual process for anything that cannot switch.

### What should an AI continuity test check?

Which workflows stop, how long the switch to a fallback takes, whether output quality on the fallback is acceptable, and who decides to switch.

---

This article is general information, not legal advice. It reflects the position as at the date of publication.

About Perfraction: a Singapore-based fractional legal function for technology companies, led by Joshua Woo. Perfraction is a legal consultancy and not a law firm. We do not provide legal advice or legal representation. Our services are limited to strategic advisory, legal operations, and regulatory consulting. For legal advice, please consult a qualified lawyer or licensed law firm.
Contact: josh@perfraction.com · Book a call: https://calendly.com/perfraction/30min · Questionnaire: https://perfraction.com/questionnaire