Call our Glasgow IT support team: 0141 255 1617

How Would You Stop AI in an Emergency?

Quick Answer: An AI emergency response plan should show where AI is being used, who owns each system and how its access, integrations or automated actions can be switched off. It should also explain how the incident will be investigated, documented and communicated.

The real question is not whether your business uses AI. It is whether you are still in control of it.

Artificial intelligence is becoming part of everyday business more quickly than many organisations realise.

It drafts emails, summarises meetings, answers customer questions and automates routine work. AI features are also quietly appearing inside software businesses already use, sometimes without anyone paying much attention to the update.

Most of the time, that is a benefit. Work becomes faster and easier.

But what happens when something goes wrong?

An AI tool sends a customer the wrong information. It exposes sensitive data. It makes an unsuitable recommendation or starts taking actions nobody approved.

Could your business stop it quickly and with confidence?

Watch this week’s Tech Update: Could you stop an AI system quickly if something went wrong?

Would you know where to start?

Ask many businesses how they would shut down an AI system during an incident, and you will often get a pause rather than a clear answer.

That is not because there is one obvious platform with a large off switch waiting to be pressed. It is because AI may be spread across several different systems, including:

  • Microsoft 365 and other cloud services
  • Customer relationship management software
  • Accounting and reporting platforms
  • Marketing and content tools
  • Customer service systems
  • Browser extensions
  • Third-party applications
  • Automated workflows and integrations

Containing a problem might mean disabling a user account, disconnecting an integration, revoking an application’s permissions, pausing an automated workflow or blocking access to a service.

Working out which action is required can waste valuable time when there is no documented plan.

You cannot stop what you cannot see

The deeper issue is that many organisations do not have a complete picture of where AI is being used.

An employee experiments with a new tool. A department activates an AI feature already built into its software. Someone connects an application to company email or cloud storage. An automated workflow is created and gradually becomes essential to a business process.

None of these changes may appear especially risky on its own. Together, they create blind spots.

If you do not know where AI is operating, you cannot properly assess what information it can access, what decisions it influences or how to contain it when something goes wrong.

Not every AI incident looks the same

An AI emergency is not always a dramatic technical failure. It could be:

  • An employee entering confidential information into an unapproved AI tool
  • A chatbot giving customers inaccurate advice
  • An automated system altering records or producing misleading financial information
  • A compromised AI-enabled account giving an attacker access to connected systems
  • An AI service sending messages or taking actions that were never properly approved

The exact response will depend on what has happened, but the first steps are usually similar.

The first six actions to take

  1. Stop the harmful activity.
  2. Prevent further access or automated actions.
  3. Preserve logs and evidence.
  4. Establish which data and systems were affected.
  5. Inform the appropriate people.
  6. Decide how normal service can be restored safely.

The National Cyber Security Centre recommends that organisations maintain an incident response plan with clearly defined responsibilities and review it regularly. Its secure AI guidance also treats monitoring, audit logging and incident response as important parts of operating AI systems securely.

This is one reason AI should form part of your wider cybersecurity planning, rather than being treated as a separate experiment.

Who is responsible when AI fails?

This is where many businesses become stuck.

If an AI tool makes a mistake, who decides whether it should be switched off? Is it IT, the software provider, the department using it or a senior manager?

It is tempting to treat AI as an IT issue by default. In reality, AI can affect operations, finance, customer service, marketing and HR.

IT may be able to disable an account or disconnect an application. It may not be able to decide whether an important business process should stop or whether customers need to be contacted.

That is why every meaningful AI system needs a named business owner, as well as technical support.

That owner should understand:

  • Why the system is being used
  • What information it can access
  • Which processes depend on it
  • Who is authorised to use it
  • What could happen if it fails
  • Who has authority to pause or disable it

Without a clear owner, decisions stall and accountability becomes unclear at exactly the wrong moment.

What should an AI emergency stop plan include?

You do not necessarily need a separate document just for AI.

For many smaller businesses, AI can be added to an existing IT, data protection or cybersecurity incident response plan.

An inventory of AI tools

List approved AI services, built-in software features, integrations and automated workflows.

Record which department uses each tool, what it is used for and what data it can access.

A named owner

Every important AI system should have someone responsible for its business use.

There should also be a technical contact who understands how access, integrations and automation can be disabled.

Clear shutdown instructions

Document the exact steps required to contain each system, rather than relying on a vague intention to sort it out.

  • Disable user accounts
  • Revoke application permissions
  • Disconnect integrations
  • Pause automated workflows
  • Rotate passwords or access keys
  • Block access from company devices
  • Contact the software provider
  • Switch temporarily to a manual process

Do not assume the person responding to an incident will remember every step while under pressure. Write the process down and keep it somewhere accessible.

Access to logs and evidence

Know where activity logs, prompts, outputs and administration records are stored.

Good records help you establish what happened, what information was affected and whether the problem has been properly contained.

A communication process

The plan should identify who needs to be informed internally and who is authorised to communicate with customers, suppliers, insurers or regulators.

Not every incident will require external reporting. However, that decision should be made calmly and based on reliable information, rather than improvised during the incident.

A safe recovery process

Switching the system back on should not be automatic.

Someone should confirm that the cause has been understood, permissions have been checked and any unsafe automation has been corrected before normal use resumes.

Could you explain what happened afterwards?

Containing the immediate incident is only part of the response.

Your business may also need to explain:

  • Which AI system was involved
  • Why it was being used
  • What data it could access
  • What went wrong
  • Who was affected
  • What action was taken
  • How a similar incident will be prevented

This becomes especially important when an AI system processes personal information or contributes to decisions affecting individuals.

The Information Commissioner’s Office expects organisations using AI with personal data to demonstrate accountability and provide appropriate transparency. Its AI documentation guidance also recommends keeping records that provide an audit trail of how AI systems are used and explained.

That does not mean every use of AI triggers a completely separate regulatory regime. It means existing responsibilities around data protection, security, fairness and accountability still apply. Using AI does not remove those responsibilities.

Ask these questions now, not during an incident

  • Do you know which systems contain AI features?
  • Do you know what company information they can access?
  • Has someone formally approved their use?
  • Does each important system have a named owner?
  • Could you pause or disconnect it quickly?
  • Would staff know who to contact if it produced something harmful or inaccurate?
  • Could you explain its role to a customer, manager or regulator?

A no or not sure to any of these questions is not a reason to stop using AI.

It is a sign of where your visibility and control need to improve.

AI should work for you, not beyond your control

AI is too useful to sideline, and it is already embedded in many of the tools businesses rely on.

The goal is not to slow its adoption. It is to make sure AI is introduced deliberately, monitored properly and backed by a plan your team can follow under pressure.

Treat important AI systems as you would any other critical business system. Know where they are, understand what they can do, decide who owns them and make sure they can be stopped when necessary.

If you are not fully sure where AI is being used across your business, Acu IT Solutions can help you review your systems, identify potential risks and put practical controls in place.

Video transcript

Read the full video transcript

Hey, it’s Annie with your latest Tech Update.

If an AI tool in your business made a serious mistake, could you stop it immediately?

AI is built into tools your team uses every day.

But in many businesses, it has been added without clear oversight.

No one really knows where it is running or how it is being used.

If something goes wrong, that is a problem.

You cannot control what you cannot see.

And if AI makes the wrong call, who takes responsibility?

The smart move is to treat AI like any other critical system, with clear visibility and control.

Ready to take control of the AI in your business? Get in touch.

That is your Tech Update. More next week.

Not Sure Where AI Is Being Used in Your Business?

We can help you review your systems, identify potential risks and put practical controls in place before an incident happens.

Speak to Our IT Experts

AI Emergency Checklist

Your business should be able to answer four basic questions:

  • Where is AI being used?
  • What information can it access?
  • Who is responsible for it?
  • How can it be stopped?

Related IT Strategy Articles

Why AI Projects Stall and How to Keep Them Moving

Why successful AI adoption depends on clear goals, ownership and business planning.

Tech Mistakes Law Firms Can’t Afford to Ignore

Common technology risks that can cause disruption, security problems and unnecessary costs.

Is Your Vet Practice’s Technology Working as Hard as You Are?

A practical guide to reviewing technology, security and continuity in a busy veterinary practice.

View all IT Strategy & Planning articles →

Related Services

Cybersecurity Services
Protect your systems, accounts and business data.

Managed IT Services
Proactive support and oversight for your technology.

Microsoft 365 Support
Secure management of Microsoft 365, Teams and cloud access.

Cloud Backup
Protect important business information and improve recovery.

Need Help Reviewing Your AI Risks?

Acu IT Solutions can help you identify where AI is being used and tighten the controls around it.

Get in Touch